text
stringlengths 1
6.27k
| id
int64 0
3.07M
| raw_id
stringlengths 2
9
| shard_id
int64 0
0
| num_shards
int64 16
16
|
|---|---|---|---|---|
Modeling the COVID-19 Dynamics and Explaining the Portuguese CaseA Contribution for Fiting Differential Equation Solutions to Epidemiological Data Compartmental epidemiological models are, by far, the most popular in the study of dynamics related with infectious diseases. It is, therefore, not surprising that they are frequently used to study the current COVID-19 pandemic. Taking advantage of the real-time availability of COVID-19 related data, we perform a compartmental model fitting analysis of the portuguese case, using an online open-access platform with the integrated capability of solving systems of differential equations. This analysis enabled the data-driven validation of the used model and was the basis for robust projections of different future scenarios, namely, increasing the detected infected population, reopening schools at different moments, allowing Easter celebrations to take place and population vaccination. The method presented in this work can easily be used to perform the non-trivial task of simultaneously fitting differential equation solutions to different epidemiological data sets, regardless of the model or country that might be considered in the analysis. Introduction Back on January 2020, when the World Health Organization (WHO) first mentioned a cluster of pneumonia cases in Wuhan, no one expected that by the 11th of March, would have spread across the globe and would be declared a pandemic [1]. Infectious diseases have accompanied mankind since the beginning of its existence and are known to have profoundly influenced the fate of entire nations upon their evolution to local epidemics or to great pandemics [2]. COVID-19, caused by SARS-CoV-2 (Severe Acute Respiratory Syndrome CoronaVirus 2), is, however, | 3,069,100 | 233771847 | 0 | 16 |
the first pandemic to ever reach every country in the world within this era of global information. Furthermore, this happened despite what might be considered the largest lock-down in history, applied by many countries. The well-known works of W. O. Kermack and A. G. McKendrick [3], published in 1927, on the compartmental SIR model, have paved the way for epidemiologists to mathematically describe the dynamics of infectious diseases and became increasingly popular among the scientific community, particularly for the analysis of the current pandemic [4][5][6][7][8][9][10][11][12]. These compartmental SIRmodels are composed of systems of differential equations having no analytical solutions. Probably, this is the main reason why only a very limited number of scientific works actually provide the fit of the SIR model to the COVID-19 related data [6,7,11] and why others prefer to fit the data using approximated analytical expressions, such as the logistic function [13,14]. The number of daily new cases, deceased and recovered related to COVID-19 is made available to the general public in close to real time [15,16], which makes it possible to test models by actually fitting the existing the existing data. Here we propose a model based on the comprehensive PSEIRD(S) model by Beira et al. [11] to fit the data sets related to the infected, deceased and hospitalized cases reported for Portugal, a country that had a severe post-Christmas outbreak. As fitting the solutions of differential equations is a such a non-trivial process and requires an extensive overhead work to implement dedicated software, an online open-access platform was developed. This | 3,069,101 | 233771847 | 0 | 16 |
platform was used to test the modified PSEIRD(S) model and may be used to test other models/assumptions. It is important to note that, as it was observed for the original PSEIRD(S) model when used to analyze the data for 11 countries, the method here presented will eventually work for any country and may be used in the validation of any model. 1 Compartment diagram of the alternative SEIRD(s) model, modified in order to accommodate protected individuals, P, to discriminate between detected, I d , and nondetected, I , infectious cases, and to include the process of vaccination. The flows between compartments are normalized and mathematically related to the model parameters, which are listed and briefly described in the box on the right hand side Model The model considers that when the virus first reaches a country the initial number of susceptible individuals is equal to the entire population of that country. As people become aware of the presence of the virus they tend to protect themselves and, as a consequence, become a part of the P compartment. The rate of this transition was considered to be proportional to the number of detected active infected and to a factor α that we decided to call the protection factor. This flow happens as a consequence of individuals following sanitary and social distancing measures, either imposed by the government or self-imposed by each individual as a result of risk perception. There is also an involuntary flux from S to P corresponding to the part of the population that is | 3,069,102 | 233771847 | 0 | 16 |
physically or geographically isolated from the virus. Multiple outbreaks can be explained by considering flows from P to S with rates k P S , which may differ over time and for specific outbreaks. These flows are a consequence of governments and/or individuals relaxing some or all of the sanitary and social distancing measures. From the susceptible compartment, individuals evolve, upon contacting with an infectious individual from compartment I , to an exposed state, E , where they are infected but cannot yet infectious. This transition happens at a rate k, which is the product of the number of daily contacts and the probability of a contact generating an infection. An exposed individual evolves to I with a characteristic time, τ I , related with the incubation period of the virus. From I a fraction d is detected and transitions into I d with a characteristic time τ Id and the remaining individuals recover and go into R with the characteristic recovery time τ R . From compartment I d , a fraction l of I d dies and becomes part of compartment D while fraction p1¡lq recovers with characteristic time τ R . Finally, vaccination and loss of immunity can also be accounted for by knowing the daily vaccinated, v, and applying the characteristic reinfection rate, k RP , in the transition from R to P, respectively. The main difference of this modified PSEIRD(S) model with respect to the original PSEIRD(S) [11] is the elimination of the infection pathway by which a fraction of the | 3,069,103 | 233771847 | 0 | 16 |
infectious individuals would not infect anyone else. Additionally, it was further assumed that the deceased came exclusively from the detected infected compartment, I d , as opposed to what was considered for the original PSEIRD(S) model. These modifications do not disregard the importance of human action in the dynamics of COVID-19, as testing (parameters d and τ Id ), risk-perception (α) and sanitary/social distancing measures (P compartment) are still considered in the modified PSEIRD(S) model and vaccination was even added to this new version. The mathematical description of the seven compartments composing this model requires the following set of differential equations: In order to fit the data for the cumulative number of detected cases, N T d , only the flow that goes into I d (see eq. 5), needs to be considered. The daily detected cases are obtained from the derivative of the total detected cases The same can be done for the daily deceased, as described by equation 10. The rate of change in the total number of infections (detected and nondetected) considers the positive flow into compartment I (see eq. 4), It is also possible to know how the number of hospitalized cases and the number of patients in intensive care units (ICU) will evolve. Both these groups are a fraction of the infected detected population, present a specific rate of hospitalization and leave this state with a characteristic time. The differential equations that characterize the flow into and from these compartments are k h and k ICU are parameters that include the | 3,069,104 | 233771847 | 0 | 16 |
percentage of active infected cases that evolve into hospitalized or ICU, respectively, and the rate at which these transitions happen. τ R h and τ R ICU are the characteristic times required for a patient to move out of the hospitalized and ICU states, respectively. Data and Model Fitting In order to perform the present study, the data sets corresponding to daily infected, total deceased, hospitalized and ICU patients in Portugal were obtained from the portuguese Directorate-General of Health website [17]. Additionally, the sets corresponding the total detected cases and daily deaths were also analyzed even though they are calculated from the daily infected and total deceased data sets, respectively. The simultaneous fit of instantaneous and cumulative data proved to contribute to stabilizing the minimization process. The model fits were performed using an open-access user-friendly online platform fitteia R [18] (fitteia.org) that uses the non-linear least-squares minimization method with a global minimum target, provided by the numerical routine MINUIT from the CERN library [19]. As the compartmental models require the resolution of differential equations, the Runge-Kutta method integrated in fitteia R was used in the minimization process. After setting the values for each population compartment and parameter at initial time, t 0 , a Runge-Kutta iteration with a time resolution fixed to 0.01 days was applied. Next, the solutions for each equation were compared with the existing data in a loop until a global least-squares minimum in the model parameters space was obtained. Fixed Parameters and General Assumptions In view of the fact that the available | 3,069,105 | 233771847 | 0 | 16 |
data sets are not enough to make an independent estimation of all model fitting parameters, assumptions had to be made for the parameters presented in Tab. 1. Although the first COVID-19 cases in Portugal were detected on March 2, 2020, (day 62) those patients were known to have developed symptoms as early as February 26 [20]. Therefore, t 0 was set to 56, which is the day of year 2020 corresponding to February 26. To accommodate for the initial growth of COVID-19 detected cases, a value of I 0 equal to one was insufficient. I 0 was, therefore, set to three, the smallest value that could explain the observed data. This indicates that on February 26 there were three infectious individuals in Portugal. The need for more than one patient zero may be related to the dissemination of the disease across different regions of the Portuguese territory. The initial value for the susceptible population, S 0 was set to the total population of Portugal, as explained in subsection 2.1. The characteristic evolution time from exposed to infectious, τ I , was set to 3.5 days. The reason for this lies in the fact that individuals become infectious about two days before becoming symptomatic. Since the incubation period for this virus is 5 days in more than 90% of the cases and outliers usually fall over this 5 day period, 3.5 seems to be a reasonable assumption [21]. The detection ratio, d, and the characteristic detection time, τ Id are variables that are interconnected. It is possible | 3,069,106 | 233771847 | 0 | 16 |
to obtain the same result if both parameters are multiplied or divided by the same value. Here we considered a combination of d=0.25 (25% of cases are detected) and τ Id =6.25 to reflect the portuguese case. As it can be observed in Fig. 2, the post-Christmas outbreak was associated with a 10% increase in the percentage of positive PCR tests. As a consequence, we reduced the fraction of detection to 0.15 for the time range associated with increased positivity (between day 358 and 399). The fraction of people that die was calculated by dividing the most recent total number of deaths by the sum of the total number of deaths and recovered. Portugal was one of the countries that did not provide systematic data for the recovered. This compartment is the one that is most subject to reporting delays and, if the reporting delays are not systematic, the data present several discontinuities that are difficult to incorporate in the model. We chose to attribute 14 days to this characteristic recovery time, based on the what was observed for other countries [11]. Model Fits The results obtained from fitting the infectious, dead and hospitalized data until February 24 are presented in Fig. 3 and the optimized parameters are summarized in Tab. 2. For all plots, the horizontal axis presents the elapsed time from January 1, 2020. As it can be observed in Fig. 3, the simultaneous fit of the modified PSEIRD(S) model to the provides very good results, which may be extended in order to obtain | 3,069,107 | 233771847 | 0 | 16 |
the projections, assuming, of course, that none of the parameters change. If confinement conditions are maintained, it is is possible to observe that the number of detected daily infections decreases to zero by May 15 (day 500) . As it can be observed, the modified PSEIRD(S) model explains complex disease dynamics, such as multiple outbreaks. These outbreaks are described as a flow of people from the P to the S compartment, whose magnitude increases for larger outbreaks, as seen from the different k P S values, presented in Tab. 2. The change observed for the k value in the beginning of September, can be assigned to the establishment of strict measures related with extending the use of masks in outdoor public places, in order to prepare a secure reopening of services, such as schools. For the hospitalized cases, it was not possible to make a fit considering a unique value of k h or k ICU . The values were different for different outbreaks but can be estimated by the initial increase observed for the different waves of infection. The different values of k h and k ICU are consistent with the fact that the rate at which patients are admitted to hospitals is largely dependent on the age stratification related with a specific outbreak (for example, an outbreak where the elderly are the most affected should be related with a larger value of k h and k ICU ) and/or to the season at which it happens (winter season should increase the infected percentage and/or | 3,069,108 | 233771847 | 0 | 16 |
the rate at which patients are hospitalized). Regarding the characteristic recovery time from a hospitalization state, it was possible to make the fits assuming a single value for the different outbreaks except for the most recent one. This outbreak was the most severe in Portugal so far and lead many hospitals to exceed their capacity for receiving COVID-19 patients. Under such conditions, patients tend to leave the hospitalized state sooner because it became increasingly difficult to keep them alive. The characteristic death time also presents different values for different outbreaks. Naturally, when the number of hospitalized increases to values close to the limits of the health care system (less than 2000 ICU beds in Portugal [22]), mortality tends to happen at a faster rate. At the beginning of the epidemic in Portugal the characteristic death time may have had a smaller value because the knowledge of the virus and of effective therapeutics was probably insufficient at that time. Remaining Compartments and Model Simulations Starting from the assumptions presented in Sect. 2.3 and the model fitting parameters presented in Tab. 2 it was possible to generate the time evolution of the compartments for which there is no available data. It is important to note that these time-series are a necessary step to solve and fit the solutions of the set of differentials equations (eq. 1-7), as it is not possible to determine one compartment without considering the remaining. The time evolution of the P, S , R, E , I and I d compartments, as well as | 3,069,109 | 233771847 | 0 | 16 |
NT d , is presented in Fig. 4. Among all the compartments, the infected non-isolated, I (Fig. 4 c)), is crucial for the description of recurrent outbursts, as these individuals, unaware of their infectious state, silently spread the disease. In fact, the dramatic outcome caused by allowing people to move freely on Christmas can be explained by the following simple reasoning based on this compartment. According to the obtained results, on Christmas eve there were about 58000 individuals in compartment I . Considering that all of these spent Christmas with four other relatives, that means at least 232000 non-isolated infected cases by new year's eve. This value might even be underestimated, as it does not account for the subsequent gatherings that might have taken place during this season festivities (e.g. multiple meetings with parents and in-laws, parties for exchanging gifts, etc). In view of this, it becomes clear that the number of daily cases is not enough to characterize the state of the pandemic. A closer analysis of the curves presented in Figs. 3 and 4 shows that, although the detected daily new cases drops to zero by May 15 (day 500), the number of infected non-isolated is around 20 on the same date. This number is more than sufficient to generate new outbreaks upon careless reopening. The model fitting software -fitteia R -also provides a straightforward way of making simulations exploring the effects of varying any model parameter in anticipation of confinement/reopening policy changes. Therefore, simulations were made in order to understand how testing can | 3,069,110 | 233771847 | 0 | 16 |
be used to decrease the present confinement duration. For instance, the level of testing is here related with d and τ Id , so we have simulated different epidemic scenarios after March 1 (day 426). The results for these simulations are presented in Fig. 5. In Tab. 3 are presented the dates and N I d values corresponding to different I -related milestones: 50 and 1 cases per 100k inhabitants. As it is economically unsustainable to maintain a state of "endless" (e.g. until I goes to zero) confinement, it is important to evaluate the secure reopening of schools and of other services,especially in view of the dramatic decrease in detected infections and deaths that is currently being observed. As a reference we considered a value of 50/100k silent spreaders, as this was the value obtained for August 31 2020, prior to the first reopening of schools. Looking at Fig.5 and Tab. 3, it becomes clear that the larger the value of d, the sooner the 50/100k plateau is reached. However, upon tripling d, it is only possible to reach the 50/100k plateau 5 days sooner, which is a relatively small gain compared to the investment that would have to be made to target such a high detection fraction. Massive testing is indeed an alternative to lock-down as pointed out in the original PSEIRD(S) work [11], but testing cannot be made in an arbitrarily slow manner. In fact, the best projected scenario corresponds to doubling the value of d while decreasing the characteristic detection time, τ Id | 3,069,111 | 233771847 | 0 | 16 |
, to two days. It is also important to stress that, as the target plateau gets lower (e.g. 1/100k), the gains that may happen as a result of changing d or τ Id become more significant. In order to test the model sensitivity to the onset of possible reopening strategies, such as reopening schools or allowing Easter celebrations to take place, the P-S flow associated with the largest increase of cases while schools were functioning (k P S =1.6) was applied either from March 1 or from April 1. Furthermore, schools reopening was paired with Easter celebrations by assuming the same P-S flow obtained for Christmas (k P S =27.8). Two cases were considered for Easter: either the P-S leakage lasts for 25 days, as was observed for Christmas, or it lasts 50 days. The results of these simulations are presented in Fig. 6. The model is clearly sensitive to the onset of school activities, especially when an Easter related leakage is considered. If schools are reopened in April, there is a smaller tendency for uncontrollable infections growth, as naturally expected. This fact, although intuitively trivial, is hardly ever expressed by actual numbers, which demonstrates the advantage of using a differential equations solver/fitter and compartmental models such as PSEIRD(S) to project different scenarios. It is important to stress that the magnitude of the leakage or its duration can have unpredictable values, depending on the efficacy of applied measures and on the extent of compliance. In case of doubling the time range within which the Easter leakage | 3,069,112 | 233771847 | 0 | 16 |
is active, the resulting epidemic wave reaches a much higher amplitude, as can be seen in Fig 6 (violet lines). If we now go back to the simulation for the P compartment presented in Fig.7 it is clear that, even taking the Christmas leakage into account, the amount of individuals that are still in this compartment is much greater than the number of people that were infected with COVID-19 in Portugal so far. This fact makes it possible to project future epidemic scenarios that are much worse than the one observed immediately after Christmas. The way to dramatically reduce this risk is to vaccinate as many individuals as possible to deplete the P compartment of people and reduce its capacity to feed the susceptible compartment. In Fig. 7 we present a few scenarios to explore the effect of different vaccination rates starting from the worst scenarios presented in Fig. 6 (violet curves). In the simulations presented in Fig. 7 it was assumed that vaccination started from March 1 with different numbers of effective daily vaccinated. In view of the fact that some vaccines require two doses to fully immunize an individual, in order to achieve the effective daily vaccinated number considered for the simulations in Fig. 6, the number of vaccine doses administered per day has to be adapted according to the type of vaccine. As it can be observed in Fig. 6, if the number of effective daily vaccinated is 30000 (60000 vaccine doses for two-dose vaccines), then it is possible to decrease the number | 3,069,113 | 233771847 | 0 | 16 |
of infectious cases to zero before the end of 2021. It is also possible to account for vaccination related immunity loss. For example, assuming that immunity lasts five months, five months from March 1, the number of individuals that will have lost immunity is equal to the number of daily vaccinated. In Fig. 8, it is possible to see the simulations based again on the worst scenario presented in Fig. 6 and where the effects of different daily vaccinated paired with a five months lasting immunity are described. As immunity loss is considered, the daily vaccinated needed to eliminate COVID-19 infectious cases by the end of 2021 increases from 30000 to 38000. Conclusion In this work we have presented the analysis of the COVID-19 epidemic data for Portugal using the modified PSEIRD(S) model and an open-access online platform -fitteia R at fitteia.org -that, among other functionalities, enables the fitting of differential equation solutions to different sets of epidemiological data. The simultaneous analysis of the infected, deceased and hospitalized data sets, along with considering both daily and cumulative data, facilitated the process of finding the set of parameters corresponding to the global leastsquares minimum. Furthermore, considering the data from the beginning of the epidemic in Portugal has put into evidence similarities, differences and variability of model parameter for the successive outbursts. The modified PSEIRD(S) model includes some aspects of social behavior and is, therefore, particularly interesting for projecting different scenarios by changing parameters that can be translated into specific actions, such as increasing the fraction of detected | 3,069,114 | 233771847 | 0 | 16 |
cases, decreasing the characteristic detection time and analyzing different numbers of daily vaccinated individuals. The attempt to predict any future evolution of the pandemic is prone to failures, as human and social behavior are essentially unpredictable. The different scenarios presented in this work illustrate how the time series of the different population compartments will evolve assuming that the general response to reopening will follow the same trends observed so far. Naturally, much worse scenarios could be evaluated, but we expect that this work will contribute to put into evidence how to avoid past mistakes. Regarding the model and the data analysis methods followed, we expect to have made clear that incorporating new compartments, parameters and assumptions (ex: different variants of the virus or age stratification) will be within reach with a relatively small effort, taking into account the technical possibilities the open-access platform at fitteia.org available for general use. We trust that both the method and the analytical possibilities offered by fitteia R could provide a positive contribution for the analysis of the current pandemic in other countries and of other epidemics. This tool provides a userfriendly way of fitting compartmental models to epidemiological data and, therefore, enables the user to make more robust projections starting from the description of past observations. In fact, if a model does not explain the details of past, there is a good chance that it will be appropriate to explore the near future. Funding This study was funded by FCT -Fundação para a Ciência e a Tecnologia. | 3,069,115 | 233771847 | 0 | 16 |
SOME VOLATILE METABOLITES PRODUCED BY THE ANTIFUNGAL-TRICHODERMA ASPERELLUM UZ-A4 MICROMYCETE , INTRODUCTION According to many scientists, more than 1.5 million microscopic fungal species are spread worldwide. However, approximately 10% of the fungi and 1% of them were studied on the spectra of secondary metabolites (Weber et al., 2007). It is known that the largest amount of natural preparation, 45% of antibiotics obtained based on secondary metabolites are produced by fungi. In this case, the share of asidialmacromycetes is 11%, while the share of micromycetes belonging to the Penicilium, Aspergillus, Trichoderma and Tolypocladium is 33%. The representatives of make up almost 99% of metabolites used in medicine and agriculture (Zhu et al., 2011). Genus Trichoderma the secondary metabolites formed are important for agriculture some of which are noteworthy for their antifungal properties against phytopathogenic fungi (Daoubi et al., 2009). Trichoderma metabolites are chemically diverse natural compounds of relatively low molecular weight produced primarily by microorganisms and plants. The secondary metabolites are biosynthezed pathways from primary metabolites (i.e., polycetides or mevalonate pathways derived from acetyl coenzyme A or amino acids) and are synthesized by certain genes. The expression of these genes is controlled by one or more global regulators (Herbert, 1989). The secondary metabolites exhibit several biological activities related to the body's survival functions, such as competing against other micro and macro-organisms, symbiosis, and ion exchange. Trichoderma the production of the secondary metabolites in fungi is often interrelated with specific stages of morphological dynamics, during the active growth phase the metabolites increase and the number of species increases. | 3,069,116 | 257742327 | 0 | 16 |
The secondary metabolites exhibit several biological functions and play an important role in regulating interactions between organisms. The are phytotoxins (secondary metabolites produced by plants phytopathogenic microorganisms), mycotoxins (secondary metabolites produced by fungi that cause disease and death in humans and animals), pigments (metabolites that form color compounds with antioxidant activity), and antibiotics (natural microbiological resistance or secondary metabolites capable of destruction) (Keller et al., 2005;Chiang et al., 2009). Among microorganisms, fungi of the Trichoderma are one of the most powerful biological agents in use today because they produce various metabolites against pathogenic microbes (Khan et al., 2020;Ming et al., 2012). Trichoderma lives in the soil and grows saprophytically on many substrates, such as tree bark, and plant roots and affects animals (a source of protein and enzyme-rich nutrients when added to feed) and plants (growth, development, microbiological protection) (Atanasova et al., 2013;Holzlechner et al., 2017). Some of the secondary metabolites produced by Trichoderma are important as drugs and a single compound (6-pentyl-a-piron) as a food flavor. Since the discovery of gliotoxin in the early 1930s, the extraction and study of metabolites from fungi of the Trichoderma genus have begun (Weindling and Emerson, 1936). Over the years, analytical studies have isolated more than 120 secondary metabolites from Trichoderma and determined their structures as well (Sivasithamparam and Ghisalberti, 1998;Reino et al., 2008). However, it is difficult to determine the exact amount of secondary metabolites produced by Trichoderma, which can form more than 1000 compounds, depending on the characteristics of the strain, environmental conditions and the sensitivity of | 3,069,117 | 257742327 | 0 | 16 |
the detection method. In recent years, genetic and genomic studies have revealed that Trichoderma secondary metabolites form new types of metabolites, taking into account biosynthetic pathways, fungal metabolism and environmental interactions (Reithner et al., 2007). These micromycetes produce several pharmaceutical and biotechnologically important secondary metabolites, including non-ribosomal peptides, terpenoids, pyrons, indolyl compounds, peptaibols, polycetides, sideophores, volatile and non-volatile terpenes (Contreras-Cornejo et al., 2016;Vinale et al., 2008;Velázquez-Robledo et al., 2011;Müller et al., 2013). One of them is a steroidal metabolite viridin, an antifungal compound isolated from various Trichoderma sp (T. koningii, T. viride, T. virens) (Golder and Watson, 1980;Singh et al., 2005). This antibiotic secondary metabolite exhibits potent antagonism to microorganisms such as Botrytis allii, Colletotrichum lini, Fusarium caeruleum, Penicillium expansum, Aspergillus niger va Stachybotrys atra (Reino et al., 2008). After reviewing these, we set the goal of our study is to analyze the secondary volatile metabolites that form the fungus Trichoderma and their properties. MATERIALS AND METHODS Study area and Material Selection The Trichoderma sp. 4 strain was selected from the soils of a cotton field infected with phytopathogenic diseases in the Bukhara region of the Republic of Uzbekistan in October 2019. Fungi Identification The soil samples were dried in air for 4 hours and the isolation of microorganisms was carried out by the method of serial dilutions. The inoculum was incubated at 28 +30°C for 5 days. Observation of the appearance of colonies was recorded for 3 to 5 days. Colonies with symptoms of Trichoderma in Petri dishes were isolated and kept clean for further | 3,069,118 | 257742327 | 0 | 16 |
study. Isolated strains were identified by classical methods based on morphology using the relevant literature (Park et al., 2005). The isolated strains were deposited at the Institute of Microbiology of the Academy of Sciences of the Republic of Uzbekistan, where they were kept at a low temperature (4-5 °C). DNA Isolation and Purification A 200 µl of the Trichoderma sp. 4 fungus sample was taken in a 1.5 ml plastic tube. Then 200 μl of 200 mM LiOAc, 1% SDS buffer was added, vortexed thoroughly and incubated at 70 o C for 5 minutes (Arnold et al., 2011). The samples were then centrifuged at 15,000 g for 5 min. The liquid portion was taken to a new plastic tube, an equal volume of 96% ethanol was added and vortexed. For DNA precipitation, the sample was stored at -20 o C for 1 h and centrifuged at 15,000 g for 5 min. The liquid in the tube was discarded and washed in 70% ethanol. The precipitate was dissolved in 100 μl of TE buffer and detected on a 0.8% agarose gel. Add another 100 μl of TE to the DNA dissolved in 100 μl of TE. 1 μl of RNAase was added, vortexed and incubated at 37 °C for 30 min. Then, 0.1 volume (20 μl) of NaAc and 1 volume of isopropanol (220 μl) were added. The sample was stored at -20°C for 1 hour and centrifuged at maximum speed for 10 minutes.The liquid in the tube was discarded and washed in 70% ethanol. The precipitate | 3,069,119 | 257742327 | 0 | 16 |
was dissolved in 50 μl of TE buffer and detected on a 0.8% agarose gel. PCR Amplification of the ITS Fragment Universal oligonucleotide primers of the internal transcribed spacer (ITS) gene were used for PCR amplification: ITS1-(TCCGTAGGTGAACCTGCGG), and ITS4-(TCCTCCGCTTATTGATATGC) (White et al., 1990). PCR amplification of DNA samples isolated from bacterial strains was conducted in the GenPak® PCR MasterMix kit. In this case, the reaction was prepared in a total volume of 20 μl, consisting of 10 μl of Dilution, 8.2 μl of double-distilled water, 0.4 μl of primer (ITS1 and ITS4) and 1 μl of DNA samples. PCR amplification optimization initial denaturation at 94 °C for 3 minutes, denaturation at 94 °C for 40 seconds, primer annealing at 55 °C for 40 seconds, elongation at 70 °C for 90 seconds, final elongation at 70 °C for 7 minutes, + ∞ at 4 o C, repeated for 35 cycles. Amplicons were detected by electrophoresis on a 2% agarose gel stained with ethidium bromide. PCR Product Purification and Sequencing For sequencing, PCR products were cut from 2% agarose gel and purified using the QIAquick® Gel Extraction Kit manual. The amount of purified PCR products was measured in a NanoDrop device. Sequencing of the samples was performed using BigDye Terminator v.3.1 cycle sequencing kit and Applied Biosystems® Genetic Analyzers, 3130 series sequence (Thermo Fischer Scientific, USA). The sequence result of this Trichoderma sp 4 strain was aligned with the species in NCBI BLAST (http://www.ncbi.nlm.nih.gov/BLAST). A phylogenetic tree for ITSof Trichoderma sp 4 was built via MEGA-X (versión10.1.8) software | 3,069,120 | 257742327 | 0 | 16 |
(Tamura et al., 2007). Dual Culture Analysis A dual culture analysis. To determine their antagonism, the strains of Alternaria alternata, Fusarium solani and Aspergillus niger, which have a phytopathogenic effect, were studied against the strain Trichoderma asperellum Uz-A4. A block (6 mm in diameter) was taken from the antagonist and phytopathogens and placed in the CDA nutrient medium in the same way on both sides of the Petri dish (5.5 mm) oppositely. The experiment was repeated 3 times. Placed in a thermostat with a temperature of 25 °C. The radius of colony antagonism on the 7th day was measured and calculated by the following formula (Mao et al., 2020). Growth Inhibition Rate Control Colony Radius Treatment Colony Radius Fermentation T. asperellum Uz-A4 strain was grown on Mandel's agar medium (in a test tube) for 6 days, and its suspension at a concentration of 106-7 spores/ml was used as an equivalent material. Microscopic fungus modified by Mandels (Mandels et al., 1962) 3; sucrose -20; (pH 5.5)) were grown on nutrient medium in 500 ml Erlenmeyer flasks. in 250 ml of nutrient medium, on an orbital shaker (shakers IKA® KS 130) at a speed of 180 rpm, at a temperature of 28-30 °C for 14 days. Filtration On the 14th day of growth, the biomass of the T. asperellum Uz-A4 strain was isolated from the culture liquid by double filtration through filter paper (Whatman #1). The culture fluid extracted from the biomass was stored at 4 °C. Extraction The separated culture liquid was extracted 3 times in a | 3,069,121 | 257742327 | 0 | 16 |
separating funnel 3:1 in ethyl acetate (EtOAc). The extraction was repeated every 2 hours. The aqueous layer at the bottom of the separating funnel was removed. The extract with ethyl acetate was dried at a temperature of 40°C under a vacuum (in a rotary evaporator) (Stracquadanio et al., 2020). Analysis of VOCs Unknown volatiles was detected in a YL 6900 GX/MS gas chromatography-mass spectrometric detector using a YL 6900 GX/MS (Young In Chromass, Korea) equipped with a DB-5MS column (30 m × 0.25 mm inner diameter, 0.25 μm film thickness) substances were identified. Oven temperature -initial -80°C/3.0 min, heating rate -15 °C/min to 250 °C, hold -3.0 min, Helium was used as carrier gas at a flow rate of 1.0 ml/min. Evaporator temperature -280 °C, flow section -1/20, analysis time -17 min. organized the Liquid samples were injected into disparities using a 1 μl microsyringe. The temperature of the transmission line was 300 °C, the ionization voltage was -70 eV, and the ion source temperature was 230 °C. Scanning range -30-350 a.m.u. The components were identified from the mass spectra of each component after comparison with the available spectral data in the MS library NIST 2017 (Guarrasi et al., 2017). RESULTS The PCR product of the ITS part of Trichoderma asperellum Uz-A4 strain (ON534075) is about 600 bp (Figure 1), and when we compared it with the data in the NCBI BLAST database, it showed that this strain is 100% similar to about 100 species of Trichoderma asperellum. We analyzed the phylogenetic tree with the species | 3,069,122 | 257742327 | 0 | 16 |
T.hamatum, T.atroviride, T.viride, T.harzianum and Protocrea pallida as an outgroup. We named this strain (ON534075) Trichoderma asperellum Uz-A4 strain based on the molecular identification of the ITS part. Phylogenetic tree of the ITS portion of Trichoderma asperellum strain Uz-A4 (blue) (Figure 2) and related species and Protocrea pallida as an outgroup. Antagonistic properties of Trichoderma asperellum Uz-A4 strain against phytopathogenic strains of A. alternata isolated from infected wheat (Triticum aestivum), A. niger isolated from infected cucumber (Cucumis sativus L.) and F. solani isolated from bean (Phaseolus L.) were studied during the research. T. asperellum Uz-A4 strain showed 77% antagonism against the A. alternata strain in 7 days, stopped the growth of the phytopathogenic fungus, multiplied on its colonies, changed the color of the colony, and acted as a super parasite. As A. niger strain is a relatively strong phytopathogen, T. asperellum Uz-A4 strain produced 55% antagonism, 97% antagonistic zones were formed compared to F. solani strain, surrounded F. solani hyphae and growth was observed in the hyphae (Figure 3). To detect volatiles in the secondary metabolites, the T. asperellum Uz-A4 strain was first grown in liquid culture and was filtered and separated from the biomass. Mass spectral gas chromatography (GC-MS) analysis was performed on the extracted liquid culture (Figure 4). Phenylethyl alcohol, 5-Hydroxymethylfurfural, dehydroacetic acid, 1-Dodecanol, 2,4-di-tertbutyl phenol, diethyl suberate, n-hexadecanoic acid, 1hexadecanol, 2-methyl-, phthalic acid, ethyl pentadactyl ester, mono(2-ethylhexyl) phthalate, octadecanoic acid were identified compared to GC-MS library base depending on the absorption rate of some spectra of the chromatogram (Table 1). When the liquid | 3,069,123 | 257742327 | 0 | 16 |
culture of the fungus T. asperellum Uz-A4 was extracted and volatile substances were detected, the presence of the substance-phenyl ethyl alcohol was demonstrated. The chromatogram showed that the absorption rate of this substance was 6,393 minutes ( Figure 5). Although phenyl ethyl alcohol is a volatile substance by nature, it further enhances the antagonistic ability of fungi of the Trichoderma genus. Phenylethyl alcohol have been shown in studies to inhibit the growth of F. Incarnatumfrom 21,68% to 74.29% (Intana et al., 2021). 5-Hydroxymethylfurfural secondary metabolite showed an absorption rate of 7,672 minutes ( Figure 6). It is known from the literature that 5-Hydroxymethylfurfural is a furfural sugar compound and is involved in the formation of enzymes (β-glucanase, cellulose) of fungi of the Trichoderma genus. Figure 4. T. asperellum Uz-A4 fungal strain culture extract GC chromatogram; 1) Phenylethylalcohol; 2)5-Hydroxymethylfurfural; 3) Dehydroaceticacid; 4) 1-Dodecanol; 5) 2,4-Di-tert-butylphenol; 6) Diethyl suberate; 7) n-Hexadecanoic acid; 8) 1-Hexadecanol, 2-methyl; 9) Phthalic acid, ethyl pentadecyl ester; 10) Mono(2-ethylhexyl) phthalate; 11) Octadecanoic acid. Dehydroacetic acid showed an absorption rate of 9,155 minutes on the GC chromatogram (Figure 7). This substance is currently used in the storage and packaging of fruit and vegetables as well as cosmetics (Saravanakumar et al., 2018). The volatility of 1-Dodecanol was relatively high, resulting in an absorption rate of 10,094 minutes on the chromatogram (Figure 8). It is known from the literature that 1-Dodecanolis an organic product by nature and is a fatty alcohol. As metabolites of the microorganism, the fungi Aspergillus niger, Rhizopus oryzae, Aspergillus terreus, Trichoderma viridе, Aspergillus | 3,069,124 | 257742327 | 0 | 16 |
flavus micromycetes were also extracted from ethyl acetate and found to be metabolites in GC-MS (Shaikh and Mokat, 2017). 2,4-Di-tert-butylphenolGC showed an absorption rate of 10,388 minutes when analyzed( Figure 9). 2,4-Di-tertbutylphenolis a metabolite with cytotoxicity, insecticidal, nematidic activity, antagonistic properties (Zhao et al., 2020;Shobha et al., 2020;Chen and Dai, 2015). In a study on T. asperellum Uz-A4 extraction, diethylsuberate GC produced an absorption rate of 11,001 minutes in the analysis ( Figure 10). n-Hexadecanoic acidGC showed an absorption rate of 13,764 minutes when analyzed ( Figure 11).The presence of n-hexadecanoic acid (6.17%) in the analysis of the metabolite T. atroviridi GC-MS. Due to the fatty acid content, its antioxidant and antibacterial properties have been known (Saravanakumar et al., 2018). 1-Hexadecanol produced a 14,007-minute absorption rate on a 2-methyl GC chromatogram ( Figure 12). Among the secondary metabolites of Trichoderma, the antioxidant content of this substance is high (Ali, 2021). Phthalic acid, ethyl pentadecyl ester showed an absorption time of 14,588 minutes ( Figure 13). This substance is characterized by the fact that it is an active enzymatic and bioactive substance among secondary metabolites (Bahaa et al., 2019). Mono(2-ethylhexyl) phthalate volatile metabolite showed an absorption rate of 14,817 minutes when T. asperellum Uz-A4 liquid culture was extracted with ethyl acetate (Figure 14). It is known from the literature that this substance is also actively involved for antifungal properties (Yang et al., 2013). Octadecanoic acid showed an absorption rate of 15,118 minutes in GC analysis ( Figure 15). Octadecanoic acid metabolite is a metabolite involved in the | 3,069,125 | 257742327 | 0 | 16 |
formation of carbohydrate containing substances (Kaushik et al., 2020). DISCUSSION Biopreparations based on Trichoderma are used in vegetable growing, melon growing, greenhouses, horticulture, viticulture and growing various ornamental plants and trees. This micromycete protects plants from phytopathogens. Increases seed germination capacity, enhances plant growth, increases metabolism, expands leaf plate surface, and improves soil structure. In recent years, when we got acquainted with the volatile metabolites identified in Trichoderma species, it was proposed as a biological protection agent that reduces the effects of plant diseases (Sunpapao et al., 2018;Andriamialisoa et al., 2004;Wonglom et al., 2019). The biological activity of Trichoderma fungi provides an important advantage in the competition with pathogens in terms of growth and development speed, competition for habitat and food. Trichoderma develops in the hyphae of phytopathogenic fungi, and wraps and destroys the cell walls with the help of lytic enzymes. As a result, it continues its life form as a hyperparasite. In addition, along with mycoparasitism, it also produces antibiotics. The limits the activity of pathogenic microorganisms (Harman, 1996). In figure 3, we can observe the development of the T. asperellum Uz-A4 strain in the hyphae of pathogenic microorganisms after antagonistic development against the F.solani strain. The authors also noted the production of volatile and non-volatile antibiotics by Trichoderma species in the biocontrol of plants, the formation of antagonistic properties against phytopathogenic microorganisms (Ubalua and Oti, 2007). Volatile metabolites identified in our studies are metabolites that affect the antagonistic properties of the strain. In recent years, in connection with the rapid development of biotechnology, | 3,069,126 | 257742327 | 0 | 16 |
interest in microscopic fungi of the genus Trichoderma, which attract the attention of researchers in connection with their practical significance for obtaining biologically active substances, plant protection products and as an active destructor of plant polysaccharides. It is known that Trichoderma emits various metabolites: growth factors (auxins, cytokines and ethylene), organic acids, intracellular amino acids, vitamins and over 100 antibiotics (Benítez et al., 2004). "TopShield" (New York), "Trichodex" (Israel), "Sternifag SP", "Trichodermin", "SellovirdineV-G20x", "Gliokladin", "Viridin" (Russia), "Fungilex J", "Fekord-2012-S" (Belarus) and other biological drugs were developed by international scientists based on fungi belonging to the Trichoderma (Reithner et al., 2007). | 3,069,127 | 257742327 | 0 | 16 |
Higher-Dimensional Loop Algebras, Non-Abelian Extensions and p-Branes We postulate a new type of operator algebra with a non-abelian extension. This algebra generalizes the Kac--Moody algebra in string theory and the Mickelsson--Faddeev algebra in three dimensions to higher-dimensional extended objects ($p$-branes). We then construct new BRST operators, covariant derivatives and curvature tensors in the higher-dimensional generalization of loop space. Introduction Infinite-dimensional Lie algebras are of crucial importance in a number of physical applications and have been a subject of intensive study for a number of years. The most well-known examples to physicists are the Virasoro [1] and Kac-Moody (KM) algebras [2,3,4], which play fundamental roles in quantum field theories in two dimensions with applications in statistical physics as well as in string theory and 2d quantum gravity [5]. Also, in three dimensions one finds the Mickelsson-Faddeev (MF) algebra [6,7,8], which arises in connection with gauge theories interacting with chiral matter (see e.g. [9,10] and refs. therein). In most situations it is convenient to interpret these infinite-dimensional algebras as extensions of simpler (classical) algebras of observables. Roughly speaking, this means that the commutation relations get augmented by an extra term; a central extension in the case of the Virasoro and KM algebras, and an abelian extension in the MF case. These extensions are of profound importance for the algebras and their representation theories, and, consequently, also for their physical applications. In all these cases the additional operators added to the algebra commute among themselves, hence the name abelian extension. In the Virasoro and KM cases they even commute | 3,069,128 | 15067079 | 0 | 16 |
with all other observables, and are therefore referred to as central. In this paper we will argue that non-abelian extensions may arise in the context of the theory of extended objects (p-branes) coupled to Yang-Mills fields. We will postulate a new algebra on the p-brane that reduces to the KM algebra for p = 1 (i.e. the string) and to the MF algebra for p = 3. For higher p's our construction gives rise to a non-abelian extension of the algebra of charge densities on the p-brane. This is of potential interest for particle physics because it has been speculated that the p = 5 case is relevant if one wants to incorporate non-perturbative effects in superstring theory by appealing to a dual formulation in terms of the five-brane [11]. Such a duality was conjectured in [12] before the five-brane was proven to exist [13]. The consequences of the existence of such a Montonen-Olive [14,15,16,17] type duality in string theory might be physically extremely important [11], involving phenomena like supersymmetry breaking and the occurrence of a non-zero potential for the dilaton. Recently [18,19,20,21,22] an attempt has been made to use a straightforward generalization of the MF algebra to higher dimensions in the context of the p-branes. In this attempt, the algebra reads [19] [T a (σ), (1.1) where, in general, the extension N is a function of the Maurer-Cartan form K on the group G pulled back to the p-brane. TheT 's generate gauge transformations under which K transforms as a gauge field. For p = | 3,069,129 | 15067079 | 0 | 16 |
1 [23], N is independent of K and the algebra in (1.1) reduces to the Kac-Moody algebra, while for p = 3 it is linear in K and one obtains the linear algebra inT and K described in [24,18]. This is the Mickelsson-Faddeev algebra familiar in the context of chiral gauge theories. For p > 3 the algebra becomes non-linear because K, as briefly explained in section 2, now appears non-linearly in N. In this paper we impose instead linearity and diffeomorphism invariance to obtain an alternative generalization that effectively extends the operator content of the theory with the introduction of new forms of intermediate degree satisfying non-trivial commutation relations among themselves, i.e., corresponding to a non-abelian extension. Thus, when generalizing many of the results known from the loop space formulation of string theory, our algebra makes it possible to avoid non-linearities even for p > 3. More specifically, we will construct a BRST operator for the p-brane wave functional, a covariant derivative in "p-loop space" and its related curvature tensor. To allow for a direct comparison with (1.1), we present here the explicit form of the algebra that we will obtain for p = 5: [T a mn (σ), T b i (σ ′ )] = 0, [T a i (σ), T b j (σ ′ )] = 0. (1. 2) The precise meaning of all the symbols appearing in (1.2) will be explained in section 3. It is important to note that K plays no role in this algebra which therefore is more general | 3,069,130 | 15067079 | 0 | 16 |
than (1.1), since K itself is connected to a special realization based on functional derivatives on the group manifold G. In fact, when constructing (1.2) in section 3 no explicit reference is made to G apart from the fact that we use forms valued in its Lie algebra. We should also remark at this point that although we construct most quantities of relevance for the p-brane algebra (1.2), we do not present a p-brane action related to it. For the algebra (1.1), on the other hand, as will be discussed in section 2, such an action exists and does indeed require the Maurer-Cartan form K for its construction. To our knowledge this is the first time non-abelian extensions appear in physics. In this paper we will limit our considerations to p-branes, although our construction also incorporates (for p = 2) certain current algebras arising in the canonical formulation of (2+1)-dimensional non-linear σ-models [25,26,27]. We hope to return to further applications in the future. To conclude this introduction, let us briefly mention the question of representation theory. Finding unitary representations of the Mickelsson-Faddeev algebra has turned out to be an extremely difficult problem. However, recently there has been some progress along these lines in the (3+1)-dimensional case using regularization techniques based on the calculus of pseudodifferential operators [28]. In this context, we believe that our algebra, being linear and not involving directly a gauge potential, could stand a better chance of having a tractable representation theory. Also, the fact that there are now two different higher-dimensional current | 3,069,131 | 15067079 | 0 | 16 |
algebras might be of some help in understanding the difficulties that one encounters in the search for unitary representations. The plan of the paper is as follows. In section 2 we review some of the previous developments emphasizing the role played by the Maurer-Cartan one-form appearing in the σ-model action, the BRST operators and the covariant derivatives. This rather detailed account, which also explains the origin of the algebra (1.1), is included in order to facilitate the comparison to our construction which is first explained for the BRST operator in section 3, and then for the covariant derivative in section 4. In section 4 we also discuss the corresponding field strength and its Bianchi identity. Section 5 is devoted to putting our new algebra into a mathematical context, explaining how it fits into the theory of extensions in terms of exact sequences etc. Finally, in section 6 we summarize our findings and make some additional comments. Formulae for some of the quantities appearing in the descent equations are collected in appendix A, together with the explicit expressions for the BRST operators and covariant derivatives. 2 The origin of the Mickelsson-Faddeev algebra for the p-brane We consider the bosonic p-brane as being defined by an embedding of the (p + 1)-dimensional worldvolume swept out by the manifold Σ p moving in spacetime M, Σ p being the p-dimensional manifold describing the p-brane itself. We will restrict ourselves to local considerations, and at the present level of understanding we also have to content ourselves with coupling the p-brane | 3,069,132 | 15067079 | 0 | 16 |
to background fields in spacetime. In particular, we assume the existence of a background Yang-Mills field A with gauge group G, and of a background antisymmetric (p + 1)-tensor B p+1 that couple to the (p + 1)-dimensional worldvolume. The dynamics of the system will be described by introducing a p-brane wave functional Φ(x) with x : Σ p → M. Our aim is to describe the "loop space" functional algebra of operators acting on Φ, with special emphasis on the BRST operator δ and the related covariant derivative D. Throughout the paper we use the convention that δ commutes with the exterior derivative d. The BRST operator then acts on the Yang-Mills field A and its ghost ω as For the antisymmetric tensor field B p+1 and its sequence of ghosts, denoted by Λ q p+1−q (q = 1, ...p + 1 is the ghost number), one has (2.4) (This notation differs from the one used in [18]; here the ghosts are all denoted by Λ.) Furthermore, the descent equations for the Chern-Simons form ω 0 p+2 and its descendants take the form δω q p+2−q = dω q+1 p+1−q (q = 0, ...p + 1). The nilpotency of δ on the background fields B and Λ above follows straightforwardly from these equations. We must now define how δ acts on the p-brane wave functional. To this end, consider first the case p = 1 (i.e. the string). When acting on the string functional Φ we write this operator where the one-form T 1 = | 3,069,133 | 15067079 | 0 | 16 |
τ aT a (σ)dσ is a Lie algebra-valued operator on S 1 . The Lie algebra generators τ a are assumed to be hermitian, satisfying [τ a , τ b ] = if abc τ c and tr τ a τ b = δ ab . We also assume the existence of a totally symmetric tensor d abc ≡ tr({τ a , τ b }τ c ). Furthermore, the scalar densityT a (σ), the dual of T 1 , is required to satisfy a KM algebra as a consequence of imposing δ 2 Φ = 0 [23]. Let us also point out that throughout this paper the pull-back of all the background forms on the p-brane is always understood. For example, in the above mentioned case we have ω = ω(x(σ)), Λ µ being the background form in spacetime. When we turn to the case p = 3, the operator δ becomes slightly more complicated, involving a new operator T a i (σ) [18] apart from the scalar densityT b (σ) which is now the dual of a threeform T 3 . Using the language of forms and pull-backs, the expression for δ given in [18] can be written as where T 1 = τ a T a i (σ)dσ i . Enforcing δ 2 Φ = 0 in this case leads to an MF type algebra forT and T i . When considering higher-dimensional cases, there are at least two alternative ways of generalizing the algebras of operators appearing for p = 1, 3. The | 3,069,134 | 15067079 | 0 | 16 |
first such construction occurring in the literature [18,19] relies upon a specific construction of the p-brane by a Kaluza-Klein approach. As emphasized in the introduction, this method introduces certain non-linearities, and the object of the rest of this section is to explain how this comes about. In section 3 we will then present a new, alternative, generalization that makes use of a linear algebra with a non-abelian extension. Let us thus review the Kaluza-Klein approach to the construction of p-branes. This approach relies on the observation that the operator T a i can be identified with the Maurer-Cartan form K a i on the group manifold G. One can then proceed to construct a locally gauge invariant action for the p-brane of the form Here the kinetic part is (i, j = 0, 1, .., p) with σ i , γ ij (γ = detγ ij ) the worldvolume coordinates and metric, respectively, and g ij (x) the pull-back to the worldvolume of the background spacetime metric g µν (µ, ν = 0, 1, .., D − 1). The current J = dσ i J i is expressible in terms of the pull-backs of the Yang-Mills field A and the Maurer-Cartan form K on the group G as J = A − K. The last term in (2.9) is the cosmological constant which is needed to make it possible to solve for the induced metric g ij by means of its field equations. The WZW part of the p-brane action reads [24] For the string (i.e. | 3,069,135 | 15067079 | 0 | 16 |
p = 1) C 2 is easily found by recalling that in the low energy limit of the heterotic string, supergravity is coupled to super-Yang-Mills [29,30] which forces one to let B 2 transform under gauge transformations according to where λ a is an x dependent gauge parameter. This conclusion can also be seen to follow from a Kaluza-Klein procedure [31,32] in which the bosonic string is compactified on M 10 × G, where M 10 is the 10-dimensional Minkowski space and G the group manifold E 8 × E 8 or SO (32). This approach also provides us with the relation Here ω 0 3 (K) is the Chern-Simons functional with A replaced by the Maurer-Cartan form K. The explicit form of ω 0 3 (K) and some other quantities appearing in the descent equations can be found in the appendix. In fact, by appealing to the descent equations we conclude that choosing δ λ b 2 = ω 1 2 (K, λ) makes h 3 is invariant under gauge transformations. Since the kinetic part (2.9) of the action (2.8) is trivially gauge invariant for p = 1, invariance for the complete action is achieved if To get this result we have also made use of the observation that δ λ B 2 = tr(Adλ) = ω 1 2 (A, λ). Then, by setting C 2 (A, K) = tr(AK) the above expression for δ λ C 2 (A, K) becomes a direct consequence of the gauge transformation properties δ λ A = dλ + | 3,069,136 | 15067079 | 0 | 16 |
[A, λ] and δ λ K = dλ + [K, λ]. This also allows us to check the gauge covariance of the current J = A − K in a trivial manner. These quantities are easily generalized to arbitrary odd p, as explained in [24], and the action (2.8) can be verified to be gauge invariant in exactly the same way as for p = 1. Thus, once the explicit forms of ω q p+2−q for q = 0, 1, and C p+1 (A, K) are found, the invariance can be established. The latter form [24,19] can be found by adopting the technique given in [33] for deriving solutions to the descent equations. Explicit formulae for p = 3, 5 can be found in [20,21]. In the above approach, with the identification T a i = K a i , one obtains for p > 3 an extension of the commutation relations of the charge densities by a composite operator N that is non-linear in the K's: (2.14) The explicit expression for N can easily be found by considering the form of the cocycle ω 2 p (K, λ, λ ′ ) and substituting two delta functions for the gauge parameters λ and λ ′ . For p = 5 one obtains and Since the K's commute among themselves, N(σ) also has vanishing commutators with N(σ ′ ) and the K's. In the next section we will show that, by enlarging the algebra of operators in loop space, it is possible to obtain a linear algebra | 3,069,137 | 15067079 | 0 | 16 |
from which all the relevant functional operators in loop space can be constructed in a natural way. The form of this new algebra is fixed uniquely by the requirements of linearity and diffeomorphism invariance on Σ p . 3 Higher-dimensional loop algebras with non-abelian extensions and the construction of the p-brane BRST operator In this section we will introduce our new algebra in its original and most general form. The first step in this process is to rewrite the action of the BRST operator on the string and threebrane wave functionals, given in (2.5) and (2.7), respectively, in a unified and compact way that immediately generalizes to any odd p: Here R is a formal sum of Lie algebra-valued forms in spacetime of even degree pulled back to Σ p , constructed from the Yang-Mills field A and its ghost ω. Furthermore, T is a linear functional of R defined by where T on the right hand side is a formal sum of operator-valued forms on Σ p of odd degree. The integral is assumed to vanish on all forms of degree different from p. We can therefore restrict the formal sums to R = R 0 + R 2 + · · · + R p−1 and T = T p + T p−2 + · · · + T 1 , respectively, and so Note also that, since the BRST operator raises the ghost number by one, the Yang-Mills ghost ω must appear linearly in R. In particular, for p = 1, the expression | 3,069,138 | 15067079 | 0 | 16 |
(2.5) for δΦ from section 2 is recovered by choosing Thus, in section 2 the commutation relations for the operators T p , T p−2 ,...,T 1 were considered to follow by enforcing nilpotency of the BRST operator. The idea is now to turn this argument around, and consider instead the algebra as the fundamental structure. We thus first postulate an algebra for the T 's and then ask for the expression for R that makes it possible to construct a nilpotent BRST operator. In this process we are led by the requirements of diffeomorphism invariance, which, in fact, already has led us to the expression (3.1), and, furthermore, that of agreement with the algebras previously obtained for p = 1, 3. In order to make the comparisons we will therefore start by considering these two cases. Also, when formulating the algebra, it is natural to use, instead of R, formal sums X = X 0 + X 2 + · · · X p−1 of Lie algebra-valued forms of ghost number zero. Hence, assume first that p = 1. Consider two zero-forms X and X ′ , and a one-form operator T on S 1 , all Lie algebra-valued. Then form T (X) and T (X ′ ) as in (3.2) and assume that these objects satisfy the algebra Taking X(σ ′′ ) = τ a δ(σ ′′ −σ) and X ′ (σ ′′ ) = τ b δ(σ ′′ −σ ′ ), it immediately follows that this algebra is equivalent to a Kac-Moody algebra | 3,069,139 | 15067079 | 0 | 16 |
at level k: Thus,T (σ), the dual of the one-form T 1 , is nothing but the current in terms of which the Kac-Moody algebra is usually expressed. Expanded in modes on S 1 according toT a (σ) = n J a n e inσ , it gives the algebra in its most common form: Having postulated the algebra, the next step is to impose nilpotency on the BRST operator when acting on the string wave functional Φ. From (3.1) we find where δR refers to the BRST variation of the background fields appearing in R. Using (3.8), the algebra (3.7) and the descent equation (2.4), the nilpotency equation δ 2 Φ = 0 then can be written as two separate equations, the latter one corresponding to the central part. It is easily seen that (3.9) is satisfied by R = −ω, while (3.10) gives the additional relation k = 2n. In the case of the three-brane, for which T = T 3 + T 1 and R = R 0 + R 2 , we make the observation that in the wedge product of T and {dR, dR ′ } there appears a term T 1 {dR 0 , dR ′ 0 } of degree three, i.e., T ({dR, dR ′ }) = 0. Hence, we take the generalization of the algebra (3.7) for p = 3 to include also terms of this kind: Introducing the term multiplyingk makes (3.11) equivalent to the MF algebra, as we will now show, by extracting the explicit form | 3,069,140 | 15067079 | 0 | 16 |
of the algebra. Apart from enabling the comparison this will reveal the different roles played by the two parameters k andk. For this purpose, it is convenient to return to the formulation in terms of commuting arguments X and X ′ , i.e. After decomposition this algebra reads where we have used X(σ ′′ ) = X 0 (σ ′′ ) + X 2 (σ ′′ ) with X 0 (σ ′′ ) = τ a δ 3 (σ ′′ − σ) and X 2 (σ ′′ ) = τ a δ 3 (σ ′′ − σ)dσ ′′j ∧ dσ ′′k , and the analogous expression for X ′ (σ ′′ ) with a replaced by b and σ replaced by σ ′ . In this notation, the commutation relations above correspond to the ones for , respectively. Note that while the k term in (3.5) for the string gives rise to the central extension, the corresponding term for p = 3 in (3.12) appears instead in the second commutator (3.14). Furthermore, the extension of the first commutator (3.13) has become operator-valued since it originates from the new term in (3.11) multiplyingk. Rescaling by k makes the operator T a i transform as a gauge field and turns the above algebra into the Mickelsson-Faddeev algebra. Note also that the algebra is referred to as having an abelian extension due to the last commutator (3.15) above. With the new term included in T (X) the previously derived conditions become To solve these equations we use the ansatz | 3,069,141 | 15067079 | 0 | 16 |
(Terms involving undifferentiated ω's may also be introduced but the solution will require their coefficients to vanish.) Inserted in (3.16) and (3.17) this ansatz gives ζ 1 − ζ 2 =k and (ζ 1 − ζ 2 )k + n = 0, respectively. We thus conclude that n = −kk and that where s is an arbitrary parameter. Before giving the ansatz for R in the case p = 5, we make the observation that the algebra (3.12) for T (X) (and hence also (3.11) for T (R)) generalizes immediately to any odd p without any further additions, and, consequently, so do the two conditions (3.16) and (3.17) above. The number of terms in the ansatz for R increases rapidly, and already for p = 5 it is advisable to do the algebra on a computer. The condition generalizing (3.17), may for that purpose be written in the form with the understanding that only terms of form degree p + 1 should be considered. Eq. (3.21) was obtained by requiring the form χ ≡ nω 2 p − 1 2 k trRdR inside the integral to be exact, and then using the descent equation dω 2 p = δω 1 p+1 . More precisely, (3.21) is the requirement that χ be closed, i.e. that dχ = 0. To be correct one should thus also add the condition χ = dΥ for some (p + 1)-form Υ. However, this condition turns out to be automatically fulfilled by the solutions to (3.16) and (3.20) for the cases at | 3,069,142 | 15067079 | 0 | 16 |
hand. The most general ansatz (without terms involving undifferentiated ghosts that would vanish anyway) contains ten parameters, not including the one present in the solution for p = 3: Subjected to the two nilpotency conditions this ansatz gives the following four-parameter solution for R: Having established the existence of a solution, it may be of interest to write out the explicit form of the algebra. In order to do that we first note that T (R) now involves three operatorvalued quantities, namely the five-form T 5 = τ aT a (σ)d 5 σ, the three-form T 3 , with the dual T a ij (σ), and finally the one-form T 1 . Repeating the steps described above for the cases p = 1, 3 generates the following algebraic structure: We can now explicitly see that the algebra of the scalar densitiesT a now has a non-abelian extension; the commutator (3.24) is augmented by an operatorT c ij that according to (3.27) does not commute with itself. Also, note from (3.26) that the operator T a i (properly rescaled by k) still transforms as a gauge field. Loop space covariant derivatives and non-abelian extensions In the previous section we have shown how, given our p-brane algebra, one can construct a nilpotent BRST operator acting on the p-brane wave functional and on the background fields. However, this is only one of the many ingredients needed to formulate the higherdimensional analogue of loop calculus for the p-brane. We should also specify what we mean by the connection one-form and | 3,069,143 | 15067079 | 0 | 16 |
the curvature two-form. Actually, from a logical standpoint, it would have been more appropriate to begin with the construction of these two objects and then describe the BRST operator. We have reversed the order in this paper because the construction of the BRST operator is somewhat simpler, not having to rely on the definition of differential forms in higher-dimensional loop space. It turns out that the extension of ordinary exterior calculus to these spaces is fairly straightforward, at least for our purposes. Let us start by choosing a basis in the space of oneforms, which we denote by δx µ (σ). In particular, on an arbitrary background vector field Ψ = dσ ′ ψ ν (x(σ ′ )) δ δx ν (σ ′ ) in loop space it takes the value δx µ (σ)(Ψ) = ψ µ (x(σ)). One can use this fact to construct an exterior derivative operator D in loop space, by setting D = dσ ′ δx µ (σ) δ δx µ (σ) , but it is more convenient never to use the explicit expression and instead rely on the formal definition of the exterior derivative when doing the calculations. The simplest case is when the operator D acts on a scalar function in loop space. To be specific, let us take the scalar T (R), whose exterior derivative is needed anyway and which can be regarded as a paradigm for all the following calculations. On such a function, D acts by creating a loop space one-form, which can be specified by giving | 3,069,144 | 15067079 | 0 | 16 |
its value on an arbitrary vector field such as Ψ defined above: (4.1) Here L ψ represents the Lie derivative with respect to the spacetime vector field ψ µ (x) and the argument of T is understood to be pulled back to the p-brane after the derivative has been taken. The first equality in (4.1) is simply the definition of the action of the exterior derivative on a loop space scalar. The second relation is slightly non-trivial and can be proven by the following explicit calculations. Instead of being too general, let us consider a specific example to show how the calculation goes. The generalization from this example should be straightforward. Thus, let us consider T (R) of the form The next case of interest is when D acts on a one-form ξ in loop space. Again, from the general expressions for the exterior derivative one has when the resulting two-form is evaluated on two arbitrary vectors Ψ 1 and Ψ 2 in loop space. The expression for a generic n-form ξ n is well-known from elementary differential geometry and it is given here for completeness: (The only additional case we will need in this paper is n = 2 for the Bianchi identities of the curvature tensor.) Now that we have the expression for the exterior derivative, we can look for a covariant version of it by adding to D a connection one-form. The idea is the same as for the construction of the BRST operator; we start with an ansatz in terms of background | 3,069,145 | 15067079 | 0 | 16 |
forms in spacetime. Such an ansatz must be compatible with our algebra and it must be diffeomorphism invariant. We then impose the covariance of the derivative in loop space and obtain the differential equations that the background fields must obey in spacetime. Again, we find it very convenient to work with formal sums of differential forms of various degree. Thus, let Ω ψ be a formal sum of forms of even degree and C ψ a p-form, both pulled back to the p-brane manifold Σ p and depending on the background Yang-Mills field A(x) and an arbitrary background vector ψ(x). (The dependence on the vector appears through the contraction of one of the indices by ψ µ before the pull-back. Explicit expressions for these forms are given in the appendix. Also, note that in this section we do not explicitly indicate the degree or the ghost number of a form.) We can now take as the covariant exterior derivative of the p-brane wave functional Φ the expression The left hand side of this equation has to be interpreted as the one-form in loop space that on an arbitrary vector Ψ takes the value where the symbol i ψ represents the contraction of the first index of the (p + 1)-form B by the background vector ψ, and the pull-back is, as always, understood. It will be clear from the following calculation why one must add the integral term in 4.7 to the covariant exterior derivative. Furthermore, eq. (4.7) corresponds exactly to the covariant derivative discussed previously | 3,069,146 | 15067079 | 0 | 16 |
in the literature [34,23]. We now want to find the equations (in ordinary spacetime) that the forms Ω ψ and C ψ must satisfy in order for (4.6) to be covariant. The BRST variation of (4.7) is and the commutator of T (R) with T (Ω ψ ) that we have assumed for our algebra, one can read off the "covariance equations" for the background forms in spacetime: These two equations express in a compact and coordinate-free notation the covariance of the operator D. The solutions for the three-brane and the five-brane are given in the appendix. Here, we simply want to remark that for the string case (p = 1) one obtains the known loop space results [23] R = −w, Ω ψ = i ψ A and C ψ = i ψ A A. Also notice that the variation of B in D has cancelled against the Lie derivative of Λ, leaving only a B-independent term ω 1 p+1 . We can now immediately construct the curvature tensor associated with the covariant derivative D. In the same way as for ordinary Yang-Mills theory, one can define the total curvature tensor G as the multiplicative operator such that DDΦ = G Φ. (Note that this formula will be modified if loop space is not torsion-free. This is what happens, for example, in the supersymmetric formulation of the problem [34,23].) By calculations analogous to those performed when deriving (4.11) one obtains, In deriving (4.12) we used the facts that L ψ = i ψ d + | 3,069,147 | 15067079 | 0 | 16 |
di ψ and that The total curvature G depends on both background forms A and B. However, this is not the right choice, as already has been noticed for the string. The problem is that H 0 is not invariant under BRST transformations. Instead, it changes according to We must therefore add and subtract the Chern-Simons form ω 0 p+2 to the total expression G and consider the splitting: This concludes the construction of the covariant derivative and the related curvature tensor in loop space. In the next section, after a review of the concept of (non-abelian) Lie algebra extensions, we will show how our algebra fits into this mathematical framework. General theory of extensions In this section we will be reviewing some more mathematically oriented issues related to Lie algebra extensions, with particular emphasis on non-abelian extensions. This is the "infinitesimal version" of the theory of group extensions [35,36,37] and it is of interest here because the algebra presented in this paper becomes a non-abelian extension of the algebra of the charge densities for p ≥ 5. The subcase of abelian extensions, and, in particular, the further subcase of central extensions, are well-known to physicists due to the abundance of situations in which they arise. Surprisingly, there has been very little discussion in the physics literature about the non-abelian case. Hence, it may be worth presenting a self-contained review of the subject, paying particular attention to the new aspects that might be of relevance to physics in the future. In principle, this section could be | 3,069,148 | 15067079 | 0 | 16 |
read independently from the rest of the paper, but at the end we will make contact with the work on the p-brane by using our new algebra as an explicit example. Let us then begin with the definition of an extension of a Lie algebra. Consider the following exact sequence of Lie algebras, 0 denoting the trivial algebra with one element: By exactness of this sequence we mean that all the arrows represent Lie algebra homomorphisms and that the image of any one of them coincides with the kernel of the following. In particular, Ker(i) = Im(δ) = 0 means that L 0 is embedded inL by a one-to-one homomorphism, and L = Ker(ǫ) = Im(π) means that π is onto. Less trivially, the condition Ker(π) = Im(i) means that i(L 0 ) is an ideal ofL and that L =L /i(L 0 ). In the following we will always identify L 0 and i(L 0 ), i being an embedding. If (5.1) is exact, we say thatL is an extension of the Lie algebra L by the Lie algebra L 0 . (To avoid confusion, we should mention that sometimes in the mathematical literature, the opposite statement is given, with L 0 and L interchanged in the above sentence. We will stick with the convention that is used in physics.) The above definition is the most general definition of an extension and we will use it without making further assumptions about L 0 . When the Lie algebra L 0 is non-abelian one may stress | 3,069,149 | 15067079 | 0 | 16 |
this fact by callingL a non-abelian extension. In the specific cases when L 0 is abelian (i.e., has identically vanishing Lie product: [L 0 , L 0 ] = 0),L is called an abelian extension. Furthermore, if L 0 is contained in the center ofL (i.e., [L 0 ,L ] = 0) we refer toL as a central extension. It is easy to see why the study of extensions is of interest in quantum physics. Suppose we have a classical system that yields an algebra L in its canonical formulation. It is well known that going to the quantum theory, by turning the Poisson brackets into commutators, one may have to add extra terms to solve ordering ambiguities, effectively enlarging the algebra toL . These terms, however, are proportional toh and therefore, by power counting, they must form an ideal L 0 ofL , i.e. [L 0 ,L ] ⊆ L 0 . This is exactly the statement thatL is an extension of L by L 0 . In the above discussion, we assumed the existence ofL in order to define the sequence (5.1). In physics, the standard situation is that we have the Lie algebra L and we want to study its possible extensions. It should be obvious that simply knowing L 0 is not enough to uniquely determineL . For example, given any two Lie algebras L and L 0 one could always construct their direct sum, and this is certainly not the most general case. We will show that the extra information that | 3,069,150 | 15067079 | 0 | 16 |
is needed is almost entirely encoded in a homomorphism θ from the Lie algebra L to the Lie algebra of the exterior derivations of Such derivations form a Lie algebra D(L 0 ) and it is easily seen that the adjoint algebra Adj(L 0 ) (i.e., the algebra generated by the adjoint action Ad of L 0 on itself), is an ideal of D(L 0 ). The algebra Ext D(L 0 ) of exterior derivations is then defined as the quotient of the algebra of all derivations by the adjoint algebra. In terms of exact sequences one can write this as Even the knowledge of θ does not uniquely fix the extension, but we will see towards the end that the freedom left at this point is very limited; in fact, it is essentially reduced to the choice of a central element. The more important question we must ask at this point is if it is always possible, given θ, to construct an extension. Thus, given the data L , L 0 and θ, let us try to construct the Lie algebraL . First of all, consider the set of all linear maps Ψ : L → D(L 0 ) such that Ψ α ∈ D(L 0 ) and θ(α) = P • Ψ α for all α in L . Contrary to θ, which is a homomorphism between L and Ext D(L 0 ), these functions Ψ may fail to be Lie algebra homomorphisms between L and D(L 0 ). The amount by which the | 3,069,151 | 15067079 | 0 | 16 |
map Ψ fails to be a Lie algebra homomorphism defines an element of the adjoint algebra. In other words, for any α and β in L there is an element χ(α, β) ∈ L 0 such that For a fixed Ψ, the map χ : L ∧ L → L 0 is of course not unique but it is defined up to a two-cochain C(L 0 ) being the center of L 0 , i.e., the kernel of the adjoint action Ad. Let us then take an arbitrary pair Ψ and χ as above and try to define the Lie product on the vector space L + L 0 as The Lie product will always be denoted by a square bracket; the particular Lie algebra to which it applies is always clear from its argument.) It should be emphasized that eq. (5.5) is not just a guess, but in fact the only possible form for the Lie product on the pairs. To see this, suppose just for a moment that there is an extensionL . One can then choose any cross section σ : L →L (i.e., π • σ = 1 ) and uniquely decompose any element λ ∈L as λ = σ(α) + a ≡ (α, a), where α ∈ L and a ∈ i(L 0 ) ≡ L 0 . The Lie product inL then ensures that The case n = 3 is of interest here and the cases n = 1, 2 will also be needed in a short while. Thus, ∂ω | 3,069,152 | 15067079 | 0 | 16 |
n (α 1 , · · · , α n+1 ) = = 1≤i<j≤n+1 (−1) i+j ω n ([α i , α j ], α 1 , · · · , α i−1 , α i+1 , · · · α j−1 , α j+1 , · · · , α n+1 ) + 1≤i≤n+1 (−1) i+1 Ψ α i (ω n (α 1 , · · · , α i−1 , α i+1 , · · · , α n+1 )). (5.9) This is the usual definition of a coboundary operator, and we know that ∂ 2 = 0. However, there is a further non-trivial check we must perform. In (5.9) we explicitly used the function Ψ, but at the beginning of this section we claimed that the extension is characterized by θ alone. We must therefore check that for any otherΨ corresponding to the same θ, eq. (5.9) gives the same result. This is fairly obvious if one realizes that two such Ψ's can only differ by an element of the adjoint algebra of L 0 , and that this element acts trivially on the center of L 0 itself. This shows that the cohomology of these algebras is specified by θ alone and not by the particular choices of Ψ and χ. The last check to perform is that ω defined in (5.8) is in the center of L 0 , which can be done by checking that it has vanishing Lie bracket with an arbitrary element of L 0 . Having checked | 3,069,153 | 15067079 | 0 | 16 |
all these points, we can now go back to (5.8) and see where the potential obstruction to constructing the extension can arise. Suppose that a particular choice of Ψ and χ satisfying (5.3) yields a non-zero ω in (5.8). We may then try to shift ω to zero by using the arbitrariness that we have in the choice of χ, and let χ → χ + η, η being an arbitrary two-cochain. Performing these substitutions in (5.8), one can see that ω is shifted by the coboundary of η: ω → ω + ∂η. However, ω need not be a coboundary, only a cocycle. We therefore come to the conclusion that the potential obstruction in constructingL is an element of the third Lie algebra cohomology group H 3 (L , C(L 0 )). Given the triple L , L 0 and θ one can, by using Ψ and χ, construct uniquely an element [ω] ∈ H 3 (L , C(L 0 )) that does not depend on the particular choice of Ψ and χ but only on θ. The Lie algebraL exists if and only if [ω] = 0. Note that this obstruction is never present in the theory of abelian extensions. It will also be absent for the non-abelian extensions we are considering. Its relevance to physics is not well-understood at the moment; it seems to suggest the possibility of having some new form of anomaly, of algebraic nature, when dealing with systems that admit non-abelian extensions. Before concluding this general review and applying the | 3,069,154 | 15067079 | 0 | 16 |
results to the algebras of interest in this paper, we must go back to the question left unanswered in the beginning of this section of how many inequivalent extensions there are for a fixed, obstruction-free θ. First, we need to specify what we mean by equivalent extensions; two extensionsL 1 and L 2 of L by L 0 are said to be equivalent if there is a Lie algebra isomorphism Φ :L 1 →L 2 that can be written as Φ(α, a) = (α, a + φ(α)) for some φ : L → L 0 . It should be obvious from the above discussion that two extensions corresponding to two different θ's cannot possibly be equivalent. Suppose now for a moment that the two extensionsL 1 andL 2 are equivalent, i.e., that there is a φ as above. Then, the pairs Ψ 1 , χ 1 and Ψ 2 , χ 2 , corresponding toL 1 andL 2 , are related by 10) On the other hand, suppose we are given the two extensions above and we want to find out if they are isomorphic. Since Ψ 1 and Ψ 2 correspond to the same θ, one can always find a map φ such that (5.10) is satisfied, and such a φ is defined up to a one-cochain λ : L → C(L 0 ). However, from equation (5.3) we can only infer that the quantity is a two cocycle in C(L 0 ), i.e. ∂η = 0. If we try to use the arbitrariness in | 3,069,155 | 15067079 | 0 | 16 |
the choice of φ to shift η to zero, we see that by changing φ → φ + λ we can shift η only by a coboundary η → η + ∂λ. We then come to the conclusion that the non-equivalent extensions are labelled by the elements of the second cohomology group H 2 (L , C(L 0 )). This is precisely the result familiar in the theory of central extensions. Hence, having fixed θ we only have the freedom of adding a central element. This concludes the review of the general theory. We can now formulate the construction of our algebra in the language explained above. When presented in this way it seems to be the most natural possible extension, given our previous knowledge for the low-dimensional cases. Thus, let Σ p be a p-dimensional manifold thought of as a space-like section of the (p + 1)-dimensional worldvolume spanned by the pbrane. The charge densitiesT a (σ) are scalar densities of weight one on Σ p . To be specific, we will think of a as an index in the Lie algebra u(N). We will work in the dual picture and think of L as the algebra of true scalars α : Σ p → u(N). If α ∈ L , the integral Σp αT is well-defined. We now want to extend this algebra in a way that preserves the full diffeomorphism invariance on Σ p . This suggests us to take L 0 to be a formal sum of u(N)-valued differential forms of degree | 3,069,156 | 15067079 | 0 | 16 |
higher than zero, plus, possibly, a central extension. We also want to retain the same expression for the Lie product that worked in the cases p = 1, 3, i.e., for the pairs a = (X, c X ), X being the formal sum of differential forms and c X a complex number. We thus take (wedge product always understood) Note that (5.13) also arises in the canonical formulation of (2 + 1)-dimensional non-linear σmodels [25,26,27]. This is further evidence that the above construction may be relevant for other applications. One could try a general expression of the kind X = X 1 µ dx µ + 1 2 X 2 µν dx µ ∧ dx ν + · · ·, but we can easily see that forms of odd degree cannot appear in L 0 if (5.13) has to satisfy the Jacobi identities. Hence, our ansatz for L 0 is the algebra of formal sums of differential forms of even degree j, 2 ≤ j < p, X = X 2 + X 4 + · · ·, centrally extended by the complex numbers, with Lie product (5.13). To avoid misunderstanding, let us stress that forms of degree higher than p, arising from the Lie product, are all vanishing and that the integral is also assumed to vanish on all forms of degree not equal to p. Finally, we must give the homomorphism θ. For our purposes, it is sufficient to give the map Ψ: Ψ α ((X; c X )) = ([α, X] +k[dα, | 3,069,157 | 15067079 | 0 | 16 |
dX]; k tr(αdX)). (5.14) In spite of its similarity with the Lie product (5.13), equation (5.14) is not an adjoint action since α ∈ L 0 . At this point, one can go back to our general discussion and check that no obstructions arise in the construction ofL . One can now easily see what kind of algebras one gets for different values of p. For p = 1, L coincides with the Kac-Moody algebra. Indeed, for p = 1 thek term is not present at all since the only forms allowed on S 1 are zero-and one-forms. For p = 2, and for all even p's in general, the k term is not present, since it represents the integral of a sum of odd-dimensional forms. In particular, the case p = 2 coincides with the algebra arising in the study on non-linear σ-models in 2 + 1 dimensions. For p = 3 one has the Mickelsson-Faddeev algebra, which is the abelian extension that appears both in the study of gauge theories with chiral fermions and in the study of the three-brane. As long as p ≤ 3, the algebraL is an abelian extension of the algebra of charge densities. For higher p's, however, and in particular for p = 5, the algebra L 0 becomes non-abelian and, consequently, the algebraL becomes a non-abelian extension of the algebra of charge densities. An example of a non-vanishing Lie bracket is the one between forms of degree two. It is for these values of p that our algebra | 3,069,158 | 15067079 | 0 | 16 |
differs drastically form those proposed previously in the literature. Conclusions and comments In this paper we have presented an alternative formulation of the higher-dimensional loop space operators which is based on a new algebra for the p-brane. The form of this algebra is essentially fixed by the requirements of linearity, closure, diffeomorphism invariance and the need to accommodate the extensions already present in the p = 1 and p = 3 cases. Some of the previously known material, needed mostly for comparison, was summarized in section 2. The explicit form of our algebra and the expression for the BRST operator were given in section 3. In section 4 we presented the construction of the covariant derivative and its curvature tensor. Finally, in section 5, we discussed some more mathematical issues regarding the theory of extensions. Throughout the paper we restricted ourselves to a local analysis. It would be of interest to study also the "global" properties of an algebra of this kind. In particular, it must be possible to give a meaning to the exponentiation of the algebra and study what further restrictions are imposed on the parameters k andk. As mentioned in the introduction, we do not yet have an action from which our algebra follows by canonical construction. In principle, such an action could be constructed by using the method of coadjoint orbits. It remains to be seen whether this can be done maintaining full diffeomorphism invariance on the (p + 1)-dimensional worldvolume. Finally, the generalization of this construction to superspace should not present | 3,069,159 | 15067079 | 0 | 16 |
much diffi-culty. It is particularly for this case that one expects to make connections with the dual picture of the superstring. A Appendix A.1 Anomaly formulae In this appendix we collect explicit expressions for some standard forms entering in the text. The particular expressions given here correspond to the convention of commuting exterior derivative d and BRST operator δ. However, the conversion to the opposite case is straightforward. We define the Chern-Simons form ω 0 p+2 (up to an exact form) by The BRST variation of ω 0 p+2 , in turn, defines the anomaly ω 1 p+1 by the descent equation Here (A.11) must be satisfied at each form level 0, 2, ...p − 1, whereas only the level p + 1 is to be considered in (A.12). We have solved these equations for the cases p = 1, 3, 5 by making proper ansätze and then using Mathematica software, developed specifically for the task, to perform the computations. There are no principal difficulties in solving the equations for higher values of p, although the number of terms in the ansatz grows rapidly with p. However, the expressions obtained are not very enlightening even for lower p's, and they are given here mostly for the sake of completeness. For the string and the three-brane, (A. 12) gives no conditions on R that do not already follow from (A.11). We can therefore use the solution for R 0 from the string for the three-brane, and the solutions for R 0 and R 2 from the three-brane | 3,069,160 | 15067079 | 0 | 16 |
for the five-brane case. We then find the following expressions: However, when determining R 4 we found that (A.12) fixes one of the free parameters in the expression obtained by solving (A.11) only. If one were to proceed to the seven-brane, the latter solution would be the proper one to take over from the five-brane case, since (A.12), which gives the nilpotency condition for the central term, should be imposed only at form level p + 1. Apart from the results above, the nilpotency equations also determine the central charge . (A.16) We have solved these equations for p = 1, 3, 5, using the results from the previous section for R and k. To begin with, we found that where φ p (A) is the form that enters in the anomaly ω 1 p+1 , and that we have given explicitly for p = 1, 3, 5 in section A.1. For Ω δx we have not been able to find similarly nice expressions. Below, we list the solutions corresponding to the choice R 2 = 1 2k [A, dw],in which case at least Ω 2,δx can be written in a fairly compact way: As was the case with R, the general solution for Ω δx for the three-brane was found to carry over to the five-brane, although in this case this happened in a somewhat less trivial manner. Finally, note that one of the originally four free parameters in R 4 was fixed by imposing the covariance conditions. | 3,069,161 | 15067079 | 0 | 16 |
Longterm renal outcome of biopsy-proven acute tubular necrosis and acute interstitial nephritis Background: Although emerging evidence suggest acute kidney injury (AKI) progress to chronic kidney disease (CKD), longterm renal outcome of AKI still remains unclear. Unlike glomerular diseases, AKI is usually diagnosed in the clinical context without kidney biopsies, and lack of histology might contribute to this uncertainty. Acute tubular necrosis (ATN) is the most common cause of AKI due to ischemia, toxin or sepsis. Acute interstitial nephritis (AIN), caused by drugs or autoimmune diseases is also increasingly recognized as an important cause of AKI. Methods: Among 8,769 biopsy series, 253 adults who were histologically diagnosed with ATN and AIN from 1982 to 2018 at five university hospitals were included. Demographic and pathological features that are associated with the development of end stage renal disease (ESRD) were also examined. Results: Rate of non-recovery of renal function at 6 month was significantly higher in the AIN (ATN vs AIN 49.3 vs 69.4%, p=0.007) with a 2.709-fold higher risk of non- recovery compared to ATN (95% CI: 1.203–6.470). During the mean follow up of 76.5±91.9 months, ESRD developed in 39.4% of patients with AIN, and 21.5% patients of ATN. The risk of ESRD was significantly higher in AIN (23.050; 95% CI: 2.420-219.533) and also in ATN (12.136; 95% CI, 1.186-24.235) compared to control with non-specific pathology. Older age, female gender, renal function at the time of biopsy and at 6 months, proteinuria and pathological features including interstitial inflammation and fibrosis, tubulitis, vascular lesion were significantly associated with progression | 3,069,162 | 240892845 | 0 | 16 |
to ESRD. Conclusions: Our study demonstrated that patients with biopsy proven ATN and AIN are at high risk of developing ESRD. (ATN) is the most common cause of AKI due to ischemia, toxin or sepsis. Acute interstitial nephritis (AIN), caused by drugs or autoimmune diseases is also increasingly recognized Background AKI is a common clinical syndrome with significant morbidity and mortality in hospitalized patients [1,2]. It also has been increasingly recognized as the major risk factor for the development and progression of CKD and a recent United States Renal Data System report showed that ATN with no recovery is responsible for 2-3% of the annual incidence of ESRD cases [3]. However, despite this huge clinical impact, longterm outcome of AKI still remains unclear. In the spectrum of AKI, ATN, caused by ischemia, toxins, or sepsis, is the most common cause of intrinsic AKI and characterized by patchy or diffuse denudation of renal tubular cells with loss of the brush border and intratubular obstruction with sparing of glomeruli [4]. However, despite these well characterized pathological features of ATN, it is usually diagnosed clinically without histological confirmation. In contrast, disease affecting the interstitium with infiltration of lymphocytes and eosinophils, termed acute interstitial nephritis (AIN) is usually diagnosed by kidney biopsy and is increasingly recognized as an important cause of AKI [5,6,7]. While kidney biopsy is a gold standard not only in diagnosis but also in prediction of outcome in various glomerular diseases, it is not usually performed in AKI. Kidney biopsy in AKI is usually indicated in the | 3,069,163 | 240892845 | 0 | 16 |
presence of active urinary sediment with possible diagnosis of diseases affecting glomeruli or vasculature or in cases of uncertain etiologies. The majority of AKI cases are diagnosed in the clinical context. The lack of specific therapeutic option coupled with risk of complications have also been a barrier for kidney biopsy in patients with AKI and thus, the value of histological features in predicting outcome has not been studied thoroughly. Here in this study, we compared longterm renal outcome of 116 biopsy proven ATN and 137 AIN cases. Rate of progression to ESRD during the mean follow up of 76.5±91.9 months were compared and pathological features that are associated with progression to ESRD were also deteremined. Participants Out of 8,769 native kidney biopsy series that have been obtained from five university hospitals in Korea: Korea University Anam Hospital, Korea University Guro Hospital, Seoul National University Hospital, Seoul National University Bundang Hospital, and Hallym University Kangdong Sacred Heart Hospital from January 1982 and January 2018, we first identified 290 patients who were histologically diagnosed with ATN or AIN in kidney biopsy. Thirty seven patients with combined glomerulonephritis or global sclerosis of > 50% were excluded and 116 patients with ATN and 137 patients with AIN were finally enrolled ( Figure 1). We also identified 106 patients with no specific abnormality in pathology and used as control. Renal tissue was obtained with ultrasonography-guided percutaneous gun biopsy, and the results were interpreted by a renal pathologist in each hospital. Clinical, biochemical, and medication data were obtained from the electronic medical | 3,069,164 | 240892845 | 0 | 16 |
records using the patient identification number and date of renal biopsy. These Institutional review board approved that informed consent is not necessary because this is a retrospective study. Definitions The eGFR was estimated by using Modification of Diet in Renal Disease Study equation [8]. Body mass index was calculated on the basis of weight and height (kg/m 2 ). Proteinuria was defined as protein ≥1+ on urine dipstick. Hypertension (HTN) was defined as systolic blood pressure (SBP) ≥140 mmHg, diastolic blood pressure (DBP) ≥90 mmHg, or use of antihypertensive medication. Diabetes mellitus (DM) was defined as fasting blood glucose ≥126 mg/dL, use of an oral hypoglycemic agent or insulin, or history of diabetes according to the electronic medical records. Steroid treatment was defined as use of steroids within 30 days before or after kidney biopsy. Cardiovascular disease was defined as angina, myocardial infarction, or stroke. Non-renal recovery was defined as eGFR <60 ml/min/1.73 m 2 at 6 months after biopsy. Data on ESRD and death were collected from the registry of the Korean Society of Nephrology on April 2018 and electronic medical records [9]. A progressor was defined as a patient who developed ESRD within the follow-up period. Renal pathology Methods of renal pathology evaluation were described previously [10]. All biopsies were evaluated using hematoxylin and eosin, periodic acid-Schiff, Masson trichrome, or periodic acid methenamine silver stains for light microscopy; immunofluorescence staining using antibodies against IgA, IgG, IgM, C3, C1q, and kappa and lambda light chains; and electron microscopy. We assessed the presence of tubular necrosis, | 3,069,165 | 240892845 | 0 | 16 |
tubular edema, interstitial inflammation, tubulitis as well as tubular atrophy, interstitial fibrosis and vascular lesion. Statistical analysis All analyses were performed using SPSS software (SPSS version 25.0, Chicago, IL, USA). Data were presented as means ± standard deviations for continuous variables and as percentages for categorical variables. Differences were analyzed using a chi-square test for categorical variables and analysis of variance for continuous variables. The Kaplan-Meier method was used for the survival curve, and statistical significance was calculated using the log-rank test. For multivariate logistic regression analysis and Cox proportional hazards analysis, variables were chosen using P<0.05 in univariate analysis, along with age and sex. A P value of <0.05 was considered significant Results Baseline patient characteristics Among 8,769 kidney biopsy series, ATN and AIN were found to be 1.6% and 1.9% respectively. The baseline characteristics of the study patients were collected at the time of kidney biopsy (Table 1). Patients with AIN were significantly older than those with control and ATN (p<0.001). Compared to control group, prevelance of DM, HTN was higher in both ATN/AIN group. ATN/AIN group showed significantly lower eGFR, suggesting the state of AKI and also lower serum albumin, cholesterol and hemoglobin level. The prevalence of 1+ or more dipstick proteinuria was highest in the AIN group (p = 0.003) and the use of steroid was significantly higher in patients with AIN (p<0.001). Pathologic findings in ATN and AIN Although ATN and AIN are distinct pathological entities with characteristic tubular necrosis and interstitial inflammation, there were significant overlap in several components of | 3,069,166 | 240892845 | 0 | 16 |
pathological characteristics. Tubular necrosis (88.0% vs 65.2%) and interstitial edema (57.4 vs 37.1%) were more common in the ATN (p<0.01), while interstitial inflammation (79.6 vs 99.2, P), tubulitis (6.5 vs 51.5), interstitial fibrosis (38.9 vs 64.9) and vascular lesion (31.8 vs 49.2) were more common in AIN (p<0.01). The percentages of global glomerulosclerosis or tubular atrophy were not significantly different between the groups ( Factors associated with progression ESRD in ATN and AIN We compared clinical and pathological features that are associated with progression to ESRD in patients with ATN or AIN. Progressors were significantly older (p = 0.001) and more likely to be women (P = 0.036) and treated by steroid (p<0.001). The nadir eGFR and also 6 months eGFR were significantly lower (p<0.001) and percentage of patients with dipstick proteinuria ≥2+ were higher in progressors (p = 0.002). Among the pathological findings, the presence of interstitial inflammation (87. were significantly associated with progression, while tubular necrosis or interstitial edema, tubular atrophy were not. The presence of interstitial edema and tubular necrosis was higher in non-progressor (Table 5). Discussion In this study, we demonstrated the followings; 1) substantial proportion of patients with biopsy proven ATN (21.5%) and AIN (39.4%) progressed to ESRD in longterm follow up 2) AIN showed worse renal outcome compared to ATN 3) older age, female gender and low nadir eGFR, 6 months eGFR were associated with ESRD progression and 4) pathological features including interstitial inflammation, tubulitis, interstititial fibrosis and vascular lesion were also associated with progression to ESRD regardless of causes. | 3,069,167 | 240892845 | 0 | 16 |
Although epidemiological studies have shown AKI increase the risk of CKD and/or ESRD, longterm renal outcome of AKI still remains unclear and lack of biopsy based studies in AKI might contribute to this uncertainty. ATN from ischemia, toxins or infection is the most common type of AKI and recently, AIN has become increasingly recognized as an important cause of AKI. However, the reported incidence of AIN was only 1-4.7% in all kidney biopsy series [6] and 10-27% in biopsies performed in patients with AKI [11]. Percentage of AIN out of total 8769 biopsies in our study (1.9%) was also comparable with previous report. The renal outcomes of AIN have been reported to be poor. In a singlecenter study of 133 patients with biopsy-proven AIN, 38% of patients achieved partial recovery while 14% showed no recovery at 6 months [12]. Another study of 157 patients with AIN also showed that 52.2% of patients developed CKD by 12months and ESRD developed in 9.4% of steroid-treated patients and in 34.4% of non-treated patients during a median follow-up of 20 months [7,13]. In line with these studies, our study also demonstrated poor renal outcome; 69.4% of patients did not recover their renal function defined as eGFR>60 ml/min/1.73 m 2 at 6 months and more importantly, we demonstrated that 39.4% of patients ultimately progressed to ESRD in a follow up period of 76.5±91.9 months regardless of steroid treatment. Although specific etiologies of AIN are not separately analyzed in this study, we could clearly demonstrated worse renal outcome in patients with AIN | 3,069,168 | 240892845 | 0 | 16 |
in very long term follow up. Relative risk of developing ESRD in AIN patients were 23 fold higher compared to control group. Significantly older age, more frequently combined interstitial fibrosis or vascular lesion, that are indices of chronicity, in AIN could be one factor facilitating the progressive CKD/ESRD. In contrast to AIN that the kidney biopsy is prerequisite for diagnosis, biopsy studies of ATN has been far more limited [14,15]. The vast majority of ATN is diagnosed in the clinical context with a help of traditional urinary indices with reasonable degree of accuracy. ATN has been considered to have a relatively good prognosis in terms of functional recovery. However, according to a study by Abdulkader et.al. renal outcome of biopsy proven ATN also seems to poor. They demonstrated that 11 out of 18 biopsy proven ATN patients showed only partial recovery of renal function and higher peak creatinine, longer hospital stay and tubulointerstitial lesion that is a sum of tubular necrosis, tubular atrophy, interstitial fibrosis and interstitial infiltrate were predictors of partial recovery. However the number of patients and follow up period were only 18 patients with 6 months, making a firm conclusion impossible [16]. Recently, a 6.64-fold increased risk of developing stage 4 CKD was observed in US veterans of more than 110,000 with ATN [17]. The annual incidence of ESRD attributed to ATN was estimated to be 3.5% in 2009-2010 [3]. However, that study has possibility to have included patients with other causes of AKI, because the definition of ATN used in the study | 3,069,169 | 240892845 | 0 | 16 |
was based on laboratory findings and diagnostic code of acute renal failure (ARF) or ATN only. In contrast, our study analyzed a relative large number of biopsy proven ATN patients (n = 116) with long term follow up to ESRD. In spite of lower rate of non recovery or progression to ESRD compared to AIN, patients with biopsy proven ATN still showed poor renal outcome; 49.3% did not achieve renal functional recovery defined as eGFR>60 ml/min/1.73 m 2 at 6 months and more importantly, 21.7% progressed to ESRD during 76.5±91.9 months that is significantly higher compared to control group. To the best of our knowledge, this is the largest study of longterm renal outcome of histologically confirmed native kidney ATN. Although specific indication or timing of biopsy are not clearly recorded, these data could give an important message that ATN from diverse etiologies might be contributing to increasing incidence of ESRD worldwide. Even after adjusting multiple patient factors, relative risk of progressing to ESRD was 12.136 fold higher in biopsy proven ATN patients compared to control group. However, given that the majority of patients with AKI are diagnosed and treated without biopsy, we still cannot answer to questions that who and what percentage of patients progress to CKD/ESRD. Kidney biopsy is an important tool in diagnosis and outcome prediction in glomerular diseases. Degree of interstitial fibrosis or number of crescents are well known histologic features in predicting outcome or treatment response. ATN and AIN are distinct pathological entities with predominant tubular necrosis and interstitial inflammation. However, | 3,069,170 | 240892845 | 0 | 16 |
pathological features including tubular necrosis, interstitial inflammation, edema, tubulitis and even chronicity indices such as tubular atrophy, interstitial fibrosis or vascular lesion are substantially overlapped in both entities. We assessed the value of these histological features in predicting renal outcome in both AIN and ATN. The presence of interstitial inflammation, tubulitis, interstitial fibrosis and vascular lesion in both ATN or AIN were significantly associated with progression to ESRD while tubular necrosis, interstitial edema or tubular atrophy were not. Although all these pathological features were not found to be an independent factor that can predict ESRD, the value of these pathological features in ATN or AIN as outcome predictors should be further assessed in larger series of native kidney biopsy studies. Given that insight regarding the role of kidney biopsy in AKI is expanding, it is possible that combining these pathological features with patient clinical and laboratory findings might improve the accuracy of outcome prediction. In addition, kidney biopsy in AKI may offer opportunities of finding newer insight into heterogenous pathogenesis, molecular mechanisms and newer therapeutic targets of human AKI. Despite several meaningful findings, our study also has limitations First, kidney biopsies were not reviewed by same renal pathologist. Second, the indication and timing of biopsy might have differed among clinicians and third, the etiologies of ATN or AIN were not determined. However, to our knowledge, this is the first study to show very longterm renal outcome of biopsy proven ATN and AIN and also suggest the possible usefulness of pathological findings in predicting outcome. In conclusion, | 3,069,171 | 240892845 | 0 | 16 |
our study demonstrated patients with biopsy proven AIN or ATN are at high risk of developing ESRD compared to control patients. Pathological features of interstitial inflammation, tubulitis, interstitial fibrosis or vascular lesion might be important in progression. Declarations Ethics approval and consent to participate This study was approved by the institutional review board of Korea University Anam Hospital (IRB approval number: 2018AN0063). Consent for publication Not applicable. Availability of data and material The datasets used during the current study are available from the corresponding author on reasonable request. Competing interests The authors declare that they have no competing interests. *BMI was measured in 170 patients † Urine analysis was measured in 166 patients. ‡ The use of steroid was analyzed in 183 patients. Figure 1 Inclusion criteria of study patients Figure 2 Incidence of non-renal recovery at 6 months. The incidence of non-renal recovery was lower in the ATN group than in the AIN group, at 49.3% and 69.4%, respectively, (P=0.007). Non-renal recovery was defined as eGFR <60 ml/min/1.73 m2 at 6 months. Figure 3 Incidence of end-stage renal disease. Renal survival in the ATN and AIN groups was significantly lower than that in the N-S group (P<0.001). | 3,069,172 | 240892845 | 0 | 16 |
Leukocyte subtypes and adverse clinical outcomes in patients with acute ischemic cerebrovascular events Background Our study aimed to evaluate whether the effects on adverse clinical outcomes, defined as death, recurrent stroke, and poor functional outcomes, differed by leukocyte subtype in patients with acute ischemic cerebrovascular events, including both ischemic stroke and transient ischemic attack (TIA). Methods We derived data from the Third China National Stroke Registry (CNSR-III). The counts and percentages of each leukocyte subtype were collected within the first 24 hours after admission. Enrolled patients were classified into four groups by the quartiles of each leukocyte subtype count or percentage. Hazard ratios (HRs) or odds ratios (ORs) and their 95% confidence intervals (CIs) of adverse clinical outcomes were calculated, with the lowest quartile group as the reference category. We used C statistics, integrated discrimination improvement (IDI), and the net reclassification index (NRI) to evaluate each leukocyte subtype’s incremental predictive value beyond conventional risk factors. Results A total of 14,174 patients were enrolled. Higher counts of leukocytes, neutrophils, and monocytes were associated with elevated risks of adverse clinical outcomes. In contrast, higher counts of lymphocytes and eosinophils were related to reduced risks of adverse clinical outcomes. Meanwhile, basophil counts seemed to not correlate with adverse clinical outcomes. Furthermore, there were also significant associations between the percentages of leukocyte subtypes and adverse clinical outcomes. Conclusions Leukocyte subtypes had different relationships with adverse clinical outcomes at 3-month and 1-year follow-up in patients with acute ischemic cerebrovascular events and could slightly increase the predictive value compared with the conventional | 3,069,173 | 235961335 | 0 | 16 |
model. Introduction The peripheral immune system can respond robustly to acute ischemic stroke (1). It is increasingly thought that the activation and variations in circulating immune cells' levels are among the early biological responses detected after acute ischemic stroke (2,3). In turn, the inflammation mediated by circulating immune cells has been identified as one of the key elements in the subsequent clinical course of acute ischemic stroke, and recent studies suggest that it can influence brain injury and clinical outcomes (4). One of the most notable inflammation responses after acute ischemic stroke is the leukocyte count change (5). Data consistently Similarly, our previous study also concluded that leukocyte count at admission was correlated with both short-and long-term clinical outcomes in acute ischemic stroke patients and might have a role as a poor prognostic factor (10). It is well known that leukocytes can be divided into neutrophils, lymphocytes, monocytes, eosinophils, and basophils. After ischemic stroke, the temporal changes in the levels of leukocyte subtypes are varied (4). Moreover, each leukocyte subtype has different immunological functions and contributes differently to the pathophysiology of atherosclerosis and cerebrovascular diseases (11,12). Therefore, leukocytes' effects on clinical outcomes in patients with acute ischemic stroke may vary depending on different subtypes, which has been revealed in several studies (3,12). However, generally, these studies still have some limitations, such as small sample sizes, short followup times, and less reporting on the outcome of stroke recurrence. Also, as far as we know, few studies have focused on the relationship between leukocytes and the clinical outcomes | 3,069,174 | 235961335 | 0 | 16 |
of transient ischemic attack (TIA) patients. Our study aimed to evaluate whether the effects on adverse clinical outcomes, defined as death, recurrent stroke, and poor functional outcomes, differed by leukocyte subtype in patients with acute ischemic cerebrovascular events, including both ischemic stroke and TIA. We present the following article in accordance with the STROBE reporting checklist (available at http://dx.doi.org /10.21037/atm-20-7931). Study design and population We derived data from the Third China National Stroke Registry (CNSR-III). The CNSR-III is a large-scale nationwide, multicenter, prospective clinical registry study of patients with acute ischemic cerebrovascular events who presented to hospitals between August 2015 and March 2018 in China. Details of the study design and major results have been described previously (13). Briefly, 15,166 patients were recruited consecutively from 201 hospitals who met the following criteria: (I) age older than 18 years; (II) diagnosis of ischemic stroke or TIA; (III) within 7 days from the onset of symptoms to enrollment; and (IV) informed consent from the patient or legally authorized representative. Acute ischemic stroke was diagnosed according to the WHO criteria and confirmed by MRI or brain CT. Among the enrolled patients in the CNSR-III, we excluded 992 patients without available complete blood counts on admission or who were lost to follow-up ( Figure S1). The ethics committee at Beijing Tiantan Hospital (IRB approval number: KY2015-001-01) and all study centers gave the CNSR-III study protocol ethical approval. All patients or their legal representatives provided written informed consent before being entered into the CNSR-III study. This study was conducted following the | 3,069,175 | 235961335 | 0 | 16 |
Declaration of Helsinki (as revised in 2013). Data collection and calculation Trained neurologists at each participating hospital systematically collected baseline data, including age, sex, body mass index (BMI), smoking and drinking status, medical history, National Institutes of Health Stroke Scale (NIHSS) score at admission, time from symptom onset to enrollment, and therapy, through face-to-face interviews or medical records. Fasting whole blood samples from venipuncture were collected in vacutainer tubes containing EDTA within the first 24 hours after admission and kept at room temperature. Afterward, the count of each leukocyte subtype was analyzed by an automated hematology analyzer at each participating hospital. All measurements were performed by laboratory personnel blinded to patients' clinical situations. Each leukocyte subtype percentage was calculated as the ratio of its absolute count to total leukocyte count. Outcome assessment Patients were followed up by face-to-face interviews at 3 months and contacted over the telephone at 1 year by trained research coordinators. Information including functional status and cerebrovascular events were queried at each follow-up. Any all-cause death and stroke recurrence during the follow-up periods were recorded. The fatality was either confirmed on a death certificate from the attended hospital or the local civil registry. Recurrent stroke included both ischemic and hemorrhagic stroke, which was confirmed from the treating hospital, and suspected events without hospitalization were judged by an independent endpoint judgment committee. The modified Rankin Scale (mRS) was used to assess patients' functional dependence, and poor functional outcomes were defined as 3≤ mRS ≤5. Baseline characteristics A total of 14,174 patients were included in | 3,069,176 | 235961335 | 0 | 16 |
our analysis. Table S1 showed that included and excluded patients' baseline characteristics were well balanced, except the included patients were younger and had a higher proportion of lipid metabolism disorder, lower NIHSS scores at admission, and lower proportions of hypertension and intravenous thrombolysis. The baseline characteristics of the included patients stratified according to quartiles of leukocyte counts are shown in Table 1. Compared to the patients with a lower leukocyte count, those in the higher quartile groups were more likely to be smoking males with lower NIHSS scores at admission, and a lower proportion of these patients received endovascular therapy and intravenous thrombolysis. In patients with a higher leukocyte count, atrial fibrillation and infection within 2 weeks before admission were less frequent, while lipid metabolism disorder was more frequent. Moreover, there were differences in age and time from symptom onset to enrollment among patients in the leukocyte count quartile groups. Leukocyte subtypes and adverse clinical outcomes The risks of adverse clinical outcomes at 1-year follow-up in the quartile groups of each leukocyte subtype count are shown in Figure 1 and Table S2. Higher leukocyte count and neutrophil count were obviously related to elevated risks of death, stroke recurrence, and poor functional outcomes at 1-year follow-up compared with the lowest quartile group taken as the reference. Similar associations were also found between monocyte count and both death and poor functional outcomes at 1-year follow-up. The above relationships were still significant after adjustments. Conversely, higher lymphocyte and eosinophil counts were associated with reduced risks of death, stroke recurrence, | 3,069,177 | 235961335 | 0 | 16 |
and poor functional outcomes at 1-year follow-up. However, after adjustments, lymphocyte count only had a significant association with poor functional outcomes, and the relationship between eosinophil count and death no longer reached statistical significance. There were no significant correlations between basophil count and death, stroke recurrence, or poor functional outcomes at 1-year follow-up. Multivariable-adjusted spline regression models for associations between each leukocyte subtype count and adverse clinical outcomes at 1-year follow-up are shown in Figure 2. The above relationships between each leukocyte subtype count and risks of adverse clinical outcomes were still valid at 3-month follow-up, as shown in Table S3. However, eosinophil count did not have a significant association with stroke recurrence at 3-month follow-up after adjustments. Furthermore, a higher basophil count was correlated with a reduced risk of poor functional outcomes at 3-month follow-up only before adjustments. Also, we assessed the risks of adverse clinical outcomes at 1-year follow-up in quartile groups of each leukocyte subtype percentage, which are shown in Figure 3 and Table S4. Patients with a higher neutrophil percentage showed higher incidences of death, stroke recurrence, and poor functional outcomes at 1-year follow-up. In contrast, lower risks of death, stroke recurrence, and poor functional outcomes at 1-year follow-up were found in patients with a higher lymphocyte percentage and eosinophil percentage. There was also an obvious association between higher monocyte percentage and a reduced risk of stroke recurrence at 1-year follow-up. Moreover, higher basophil percentage was observed to be correlated with lower risks of death and poor functional outcomes at 1-year followup. | 3,069,178 | 235961335 | 0 | 16 |
After adjustments, the above relationships remained significant. As shown in Table S5, the above associations between each leukocyte subtype percentage and the risks of adverse clinical outcomes were still valid at 3-month follow-up. Furthermore, higher monocyte percentage was related to a reduced risk of death at 3-month follow-up regardless of adjustments. Higher basophil percentage only had a significant association with stroke recurrence at 3-month follow-up before adjustments. Incremental predictive value of leukocyte subtypes We evaluated whether each leukocyte subtype could further increase the predictive value of conventional risk factors for adverse clinical outcomes at 1-year follow-up, shown in Table 2 and Table S6. At 1-year follow-up, the C statistic of the conventional model for death significantly improved with the addition of neutrophil count, monocyte count, or lymphocyte p e r c e n t a g e . T h e d i s c r i m i n a t o r y p o w e r a n d r i s k reclassification also appeared to be slightly better with the addition of leukocyte count, neutrophil count, lymphocyte count, monocyte count, neutrophil percentage, lymphocyte percentage, or eosinophil percentage. As for stroke recurrence at 1-year follow-up, the addition of leukocyte count, neutrophil count, neutrophil percentage, lymphocyte percentage, or eosinophil percentage could improve the C statistic of the conventional model. Furthermore, adding leukocyte count, neutrophil count, neutrophil percentage, lymphocyte percentage, or monocyte percentage could make both discriminatory power and risk reclassification better in a bit. Meanwhile, eosinophil count or eosinophil percentage might only | 3,069,179 | 235961335 | 0 | 16 |
improve the risk reclassification. However, basophil percentage might even worsen the risk reclassification. There were also improvements in the C statistic for the conventional model for poor functional outcomes at 1-year follow-up with leukocyte count, neutrophil count, monocyte count, neutrophil percentage, or lymphocyte percentage. The discriminatory power and risk reclassification might be ameliorated by adding leukocyte count, neutrophil count, lymphocyte count, monocyte count, neutrophil percentage, or lymphocyte percentage. Similarly, eosinophil count or eosinophil percentage seemed only to improve the risk reclassification. Neutrophil count, ×10 9 Neutrophil count, ×10 9 Neutrophil count, ×10 9 Lymphocyte count, ×10 9 Lymphocyte count, ×10 9 Lymphocyte count, ×10 9 Monocyte count, ×10 9 Monocyte count, ×10 9 Monocyte count, ×10 9 Eosinophil count, ×10 9 Eosinophil count, ×10 9 Eosinophil count, ×10 9 Basophil count, ×10 9 Basophil count, ×10 9 Basophil count, ×10 9 Discussion This analysis demonstrated that different leukocyte subtypes had different relationships with adverse clinical outcomes at 3-month and 1-year follow-up in patients with acute ischemic cerebrovascular events. Higher counts of leukocytes, neutrophils, and monocytes were associated with elevated risks of adverse clinical outcomes. In contrast, higher counts of lymphocytes and eosinophils were related to reduced risks of adverse clinical outcomes. Meanwhile, basophil count seemed to not correlate with adverse clinical outcomes. Besides the counts of leukocyte subtypes, there were also significant associations between the percentages of leukocyte subtypes and adverse clinical outcomes. Neutrophils are one of the most important leukocyte subtypes and are the first leukocyte subtype to upregulate expression substantially and invade injured brain | 3,069,180 | 235961335 | 0 | 16 |
tissue after acute ischemic stroke (14). It has been found in experimental models that the invasion of neutrophils could be detected as early as 5 hours after ischemic stroke onset and peak at 24-48 hours (15). Previous studies have shown that neutrophils could aggravate ischemic brain injury, and higher neutrophil counts are correlated with increased stroke severity, infarct volume, and worse functional outcomes (12,16,17). Some experimental studies have suggested that depleting circulating neutrophils, inhibiting neutrophil infiltration, or blocking the neutrophil pro-inflammatory function could reduce infarct volume and improve neurological outcomes (18,19). Furthermore, Semerano et al. reported that even after adjusting for the occurrence of infections, the association of higher neutrophil counts with worse 3-month functional outcomes was still significant (3). Beyond neutrophil counts, Zhu et al.'s study revealed that high neutrophil ratio levels were also related to increased risks of new stroke and composite events in patients with minor ischemic stroke or TIA (20). Consistent with these prior studies, we found that patients with acute ischemic cerebrovascular events with higher neutrophil counts or percentages were more likely to suffer death, stroke recurrence, and poor functional outcomes at 3-month and 1-year follow-up. Nowadays, many mechanisms for neutrophil aggravation of ischemic brain injury have been proposed, for example, blocking microvessels, interacting with platelets, and releasing deleterious substances or inflammatory mediators (14,19,21). In contrast with neutrophils, it has been found that there is an exponential decrease in the peripheral circulating lymphocyte count in the hours immediately after ischemic stroke (4). Lower lymphocyte count has been revealed to be a | 3,069,181 | 235961335 | 0 | 16 |
marker of severe brain damage and is predictive of poor neurological improvement during the first week, poor functional outcomes at 3 months, and stroke-associated infections (12,22). Our analyses demonstrated similar results in that a higher lymphocyte count was correlated with reduced risks of poor functional outcomes at 3-month and 1-year follow-up in patients with acute ischemic cerebrovascular events. Moreover, higher lymphocyte count might also be related to lower death and stroke recurrence risks at 3-month and 1-year follow-up. As for lymphocyte percentage, patients with higher lymphocyte percentages tended to have reduced incidences of death, stroke recurrence, and poor functional outcomes at 3-month and 1-year follow-up regardless of adjustments. In short, the above findings suggested that lymphocytes might play a protective role in ischemic brain areas; however, the underlying mechanisms are still not clear. One possible mechanism is that the percentages of the disease-limiting protective lymphocyte subtypes which can produce antiinflammatory cytokines and maintain immune homeostasis, such as regulatory T cells, increase in the brain after ischemic stroke (23,24). Another possible mechanism is that lymphopenia may reflect increased sympathetic tone and serum cortisol levels, which are associated with increased production of proinflammatory cytokines (12,25,26). Although monocyte infiltration within ischemic brain tissue peaks at 7 days after acute ischemic stroke onset, peripheral circulating monocyte count increases earlier (27,28). Meanwhile, it has been found that the temporal trends of different monocyte subtypes after stroke might not all be increased. For instance, the proportion of CD14 high CD16-monocytes do not significantly change, CD14 dim CD16+ monocytes decrease, and only CD14 | 3,069,182 | 235961335 | 0 | 16 |
high CD16+ monocytes increase (29). Besides that, the effects of different monocyte subtypes on prognosis after stroke also vary, as Urra et al. suggested that classic CD14 high CD16-monocytes had deleterious effects, whereas rare CD14 dim CD16+ monocytes and CD14 high CD16+ monocytes were associated with clinical benefits (29). Overall, more studies have shown that monocyte count is positively correlated with the severity of brain injury and infarct volume after acute ischemic stroke and could be an independent predictor of poor clinical outcomes in patients (4,28,30,31). Similarly, our results also demonstrated that higher monocyte count had an obvious association with elevated risks of death and poor functional outcomes at 3-month and 1-year follow-up in patients with acute ischemic cerebrovascular events; however, there was no relationship between monocyte count and stroke recurrence. In contrast, monocyte percentage had a negative association with death at 3-month follow-up and stroke recurrence at 3-month and 1-year follow-up, which might be due to the influence of leukocyte count as the denominator. The mechanism underlying the role of monocytes after ischemic stroke is complex. On the one hand, monocytes could contribute to the development of inflammation and promote thrombosis and vascular occlusion by producing inflammatory mediators and forming platelet-monocyte aggregates, which aggravate cerebral ischemic injury (27,30). On the other hand, macrophages transformed by monocytes in the ischemic area might lead to angiogenesis stimulation, which could be considered beneficial (27). Hypereosinophilia has been reported as an unusual cause of ischemic stroke due to thromboembolism from endomyocardial fibrosis or vascular endothelial toxicity of eosinophilic cells | 3,069,183 | 235961335 | 0 | 16 |
(32,33). However, in patients without previous hypereosinophilia, eosinophils might play a protective role in the same way as lymphocytes after acute ischemic stroke. Previous studies have demonstrated that eosinophil count is negatively associated with stroke severity, risk of mortality, and poor functional outcomes in acute ischemic stroke patients (3,34). It has even been shown that acute ischemic stroke patients with lower eosinophil counts are more likely to have functional impairment in limbs and difficulty in recovering (35). Moreover, higher eosinophil count at admission is suggested to be an independent predictive factor for lower odds of developing hemorrhagic transformation after treatment with recombinant tissue plasminogen activator for acute ischemic stroke (32). A similar relationship has been found between eosinophil percentage and acute ischemic stroke prognosis (34). Our results supported these previous studies in that higher eosinophil count was related to reduced risks of stroke recurrence at 1-year follow-up and poor functional outcomes at 3-month and 1-year follow-up in patients with acute ischemic cerebrovascular events. There also might be negative associations of eosinophil count with stroke recurrence at 3-month follow-up and death at 3-month and 1-year follow-up. Furthermore, lower risks of death, stroke recurrence, and poor functional outcomes at 3-month follow-up and 1-year follow-up were shown in patients with higher eosinophil percentages. There are a few hypotheses which may explain how eosinopenia leads to poor prognosis after acute ischemic stroke. Eosinophils can secrete numerous cytokines, growth factors, and chemokines to induce the activation of M2 phenotype microglia, which have neuroprotective properties and may facilitate the resolution of inflammation | 3,069,184 | 235961335 | 0 | 16 |
(32). Another mechanism may be the correlation with angiogenesis promotion, since © Annals of Translational Medicine. All rights reserved. Ann Transl Med 2021;9(9):748 | http://dx.doi.org/10.21037/atm-20-7931 eosinophils can produce vascular endothelial growth factor (32,35,36). Up to now, few studies have investigated the response of basophils after acute ischemic stroke, and data on the relationship between basophils and prognosis are also limited. Our analyses showed that there were no significant correlations between basophil count and death, stroke recurrence, or poor functional outcomes at 1-year followup in patients with acute ischemic cerebrovascular events; however, at 3-month follow-up, higher basophil count might only be associated with a reduced risk of poor functional outcomes. Basophil percentage was observed to be negatively related to death, and poor functional outcomes at 3-month and 1-year follow-up, and might also have a negative relationship with stroke recurrence at 3-month follow-up. However, we thought that this negative association between basophil percentage and the prognosis of acute ischemic cerebrovascular events was because of the leukocyte count's influence as the denominator. Based on these results, we speculated that basophils might play a limited role in developing acute ischemic cerebrovascular events due to their low content. Nevertheless, further studies are required. Compared with the conventional model, we observed that the counts or percentages of leukocytes or each subtype, except for basophils, could slightly increase the predictive value of prognosis at 1-year follow-up in patients with acute ischemic cerebrovascular events. Of the leukocyte subtypes, neutrophils and lymphocytes seemed to have better predictive value. Therefore, as valuable peripheral blood biomarkers, leukocyte | 3,069,185 | 235961335 | 0 | 16 |
subtypes require further study to unravel their underlying mechanisms after acute ischemic cerebrovascular events and the therapeutic implications that might be derived. There are several limitations to our study. Firstly, there is inevitable equipment heterogeneity across participating hospitals, which could have contributed to biased estimates. However, this might have had little impact because of daily practices and strict quality control. Secondly, our study only collected the counts and percentages of each leukocyte subtype within the first 24 hours after admission and did not evaluate the dynamic changes. Nevertheless, it might still provide valuable information on leukocyte subtypes' underlying mechanisms after acute ischemic cerebrovascular events. Thirdly, our study did not further assess whether there was a difference in each leukocyte subtype's relationship with adverse clinical outcomes between ischemic stroke patients and TIA patients. Finally, residual bias still exists due to the influence of comorbidities or environmental factors such as tumors, trauma, and acute toxicosis. Conclusions Different leukocyte subtypes had different relationships with adverse clinical outcomes at 3-month and 1-year follow-up in patients with acute ischemic cerebrovascular events. Additionally, leukocyte subtypes, except for basophils, could slightly increase the prognosis's predictive value at 1-year follow-up in patients with acute ischemic cerebrovascular events. Acknowledgments We thank all study participants, their relatives, the members of the survey teams at participating hospitals of the Third China National Stroke Registry, and the project development and management teams at the Beijing Tiantan Hospital. Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or | 3,069,186 | 235961335 | 0 | 16 |
integrity of any part of the work are appropriately investigated and resolved. Our study derived data from the CNSR-III. The ethics committee at Beijing Tiantan Hospital (IRB approval number: KY2015-001-01) and all study centers gave ethical approval of the CNSR-III study protocol. This study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). All patients or their legal representatives provided written informed consent before being entered into the CNSR-III study. | 3,069,187 | 235961335 | 0 | 16 |
Morphological Neural Pre- and Post-Processing for Slavic Languages While developing NMT systems for our customers involving Slavic languages, we encountered certain issues that do not affect Latin or Germanic languages. The most striking of these is the morphological complexity inherent in a remarkable number of unique synthetic forms. For each language combination, the aim is always to find the best balance between the size of the vocabulary, the quality of the translation and the performance of the MT model (both training time and translation time). When working with Slavic idioms, the variety of cases and genders makes this challenge even more difficult and engaging. For Slavic source languages, our solution is to add an extra pre-processing step before the actual translation, in which the inflected word is reduced to its components; naturally, in the opposite direction this requires a symmetrical post-processing technique. Tests have proven high-quality results for Slavic languages, either source or target, confirming this as an effective approach. cells: 7 cases, 4 genders, 2 numbers.Luckily, because many of them are the same, there are "only" 11 unique variants. These forms are not as frequent in a corpus: some of them may be used ten times less than others, and this can obviously cause the engine to inconsistently translate what appears to be the same word. As you can see in Table 1, there are many more Czech forms than English ones, and our engine must be able to handle all of them.What makes this task even more difficult is that the customer's training material | 3,069,188 | 202670714 | 0 | 16 |
is often extremely repetitive, with similar forms repeated many times and others just a few.Table 1.Sample of Czech inflections of adjectives and substantives. Aim When working with standard tokenization, the initial basic conditions required to achieve good MT translations are quality and the amount of training data.There are two typical scenarios: Huge, well-formed corpora that need more extensive technical resources for training (GPU, memory, RAM, etc.) Smaller data sets, from which it is often not easy to obtain high-quality results In both cases, we can improve the process by tweaking the tokenization in a way that allows for intelligent handling of inflections.This can lead to better structuring of the engine's vocabulary, resulting in a win-win situation: instead of filling it with many variants of the same word, it can be made smaller and more efficient without sacrificing quality, or it may To je pěkná kniha.This is a nice book.To jsou pěkné knihy.These are nice books.Viděl jsem tě s pěknou knihou. I have seen you with a nice book.contain more terms from different contexts without increasing its size.In simple terms, we could obtain substantial benefits if we could separate stems from affixes. Another important consideration is that our scenario involves final users with little or no knowledge of one of the languages; in this context, reducing Out Of Vocabulary (OOV) words would be a significant goal. A solution between standard tokenization and BPE With this in mind, we need a tokenizer that works not only on word boundaries, but also in terms of the morphological | 3,069,189 | 202670714 | 0 | 16 |
construction of the token.In this respect, the BPE (Byte-Pair-Encoding) algorithm (Sennrich et al., 2016) may be a valid option, but it is based on the most common sequences of characters and thus it cannot always split words in the way a human would.It is certainly practical in the absence of further grammatical information, but it has already been proven (Ataman et al., 2017) that considering morphological aspects while tokenizing results in higher translation quality.While observing the inflections in languages such as Czech or Polish, we noticed that the ending may vary depending on the final part of the stem, which means it would be too difficult to manually split the text using a complete list of endings.In addition to this, some of them would be too rare to be learned well by the engine.We therefore supposed that, since a native speaker can implicitly distinguish stems and inflections, a neural model (from now on referred to as a Morpho Model) could be trained to do the same; that is, identify the sequence of letters that can influence the ending and split the word into stem and affix before sending anything to the translation engine.The output tokens from this preprocessing model are the ones that the final translation engine will learn. This approach differs from pure characterbased neural machine translation in that the Morpho Model only needs to parse single complete words rather than translate whole sentences. Of course, this model is only the core of this pre-processing technology, and can only produce high-quality results as part of | 3,069,190 | 202670714 | 0 | 16 |
a series of steps that guarantee clean input and output data.For example we noticed in the very first phase of tests that irregular forms had to be recognized and handled separately; in fact they represent a relatively small amount of widely used lemmas, with inflections which are hard to be learnt in a general abstract way. The attempt to find a valid solution that was different from BPE came from the need to have a sort of control over the translation.With the integration of the Morpho Model, as described in the following chapters, we can minimize the risk of unexpected phaenomena, like sub-sets of words considered sequences to be inflected.For our user case it is extremely important to have an output that fulfils the customer's needs regarding not only the general quality of the translation, but also the usage/avoidance of certain forms: therefore we chose to invest resources in a system we can control under almost any aspect. Description of the method In order to successfully implement this process, it is essential to have a map with a sufficient number of examples and a good description of many morphological categories (for example, it would not be enough to know only the gender of a noun, without its case, number, etc.). The databases we used to create the maps are free online resources.To have an idea of how big the maps are that we used, we can say that our Russian map has more than two million entries, while the Polish one has more than five million.A | 3,069,191 | 202670714 | 0 | 16 |
reduction of the map's size may be possible by comparing words in the training material for the final NMT engine with the contents of the map.Nevertheless, even words which are not contained in the customer's dictionary may help build a more consistent Morpho Model; in fact it should be trained to build up inflections with their letters, regardless of their meaning or occurrences. Since we are working with Slavic languages as either the source or the target, the Morpho Model is used in both directions; that is to say, from an inflected word to its corresponding morphological information as well as in the opposite direction.To obtain the expected benefits for the engine's vocabulary, we need to train it using a corpus where all inflections have been reduced.However, we also want to be able to parse the engine's output back to a human-readable language, so the reduction needs to be mapped towards a real word. Although the two directions have the same logic (from opposite perspectives), they may present distinct challenges during the translation process, once the engine has been trained. Slavic source language Figure 1.Overview of the process when the Slavic language is the source. When translating from a Slavic language, the Morpho Model must parse an inflected word into its grammatical information so that the engine has everything it needs to translate properly.Figure 1 shows how the whole workflow should operate; when moving from one step to the next, further text handling may be needed, such as tokenizing or checking the format. Since the same | 3,069,192 | 202670714 | 0 | 16 |
inflection can be mapped with many definitions (see Table 2), we must ensure that the Morpho Model produces output that can be used by the engine to guarantee a high-quality result; an even more difficult example is that of terms which can belong to two or more different parts of speech, like substantives and verbs or adjectives and verbs.In any case, we should remember that all languages of our experience have ambiguous words which can be understood only with the help of the context and it is one of the NMT engine's tasks to find the correct translation for each of them. Slavic target language When translating into a Slavic language, the Morpho Model is employed from the definition to the inflection.In this case, the engine plays a dominant role.In fact, its translation constitutes the input for the Morpho Model, and it must be extremely reliable in order to correctly build the final word.Consequently, particular care is required when selecting the tokens to be sent to the Morpho Model (it works at word level, so it needs one stem and several properties to generate one inflected form). There is a risk of creating incorrect or even artificial words at the end of the process, but our tests show that this risk is minimal. Figure 2. Overview of the process when the Slavic language is target. Results Test results2 involving only the Morpho Model show that when the Slavic language is the source language, the percentage of perfect matches3 is around 80%.This value is perfectly respectable, considering | 3,069,193 | 202670714 | 0 | 16 |
that the remaining non-perfect matches may fall into one of three categories: Alternative definition Correct stem with a mistake in the morphological properties Mistake in the stem While the first two cases may cause a degree of confusion and lower the final BLEU evaluation, only the third one actually represents a disturbing factor when used as input for the incoming translation engine. In any case, we can observe quite astonishing results in the opposite direction (i.e.Slavic as the target language), where the perfect match rate is over 90% for Russian, and even 97% for Polish.The difference up to 100% represents cases in which the user may receive a spurious word that does not really exist, but such an outcome can be avoided or at least strongly reduced with a simple spellchecker, for example. As regards the evaluation of the whole translation process, results appear not so easy to evaluate.If we take Polish as an example (but the other languages had similar behaviour) we see that pure BLEU values with Morpho Model are in both directions lower than the BPE. 4 Since the number of translations with BLEU below 0.2 was much bigger in the Morpho case than in the BPE, we took a selection of 150 of them and let them be analysed by translators who did not know about our study.We expected to find that recurrent phaenomena showed some kind of inconsistency in one or more steps of our process, but we were told that actually the translation with the Morpho Model | 3,069,194 | 202670714 | 0 | 16 |
often had a better level of comprehensibility.As a final test we let the translators make manual comparisons of BPE and Morpho translations in our web application, with particular focus on the correctness of inflected forms.After this confirmation we decided to use this new technology in production; in fact, we usually proceed only after the approval of a translator or at least a native speaker, especially for such cases when the automatic evaluation doesn't show a significant advantage for a particular case. Possible drawbacks Some reservations have been expressed concerning the time spent on a single translation, as each word has to be handled by the Morpho Model in addition to the time required by the normal NMT engine.In this respect, it is important to note that the Morpho Model is much faster than a conventional engine due to the consistency of material and the low settings required for its training (word vector and RNN far below 100). Another criticism may be the risk of having less control over the translation, since we are using two neural models instead of one.However, thanks to other pre-/post-processing steps, we can reduce the possibility of unexpected results, as a last resort leaving the source word un-changed to prevent the model from creating spurious words. In any case, as a company, we need to consider any MT solution in a practical way: the worst possible output for our average user is an OOV.Thus, reduction of OOVs, coupled with more consistent quality when translating the same lemma, is a major objective.In most | 3,069,195 | 202670714 | 0 | 16 |
cases, a translation containing an OOV is completely incomprehensible, while one containing the correct stem and an incorrect ending is sufficient to justify continuing with the work. Furthermore, an error rate of 3%, as the one we had for Polish, is probably not far from the human one, especially considering that not everyone among our target users has high linguistic skills. You might assume that a technique based on morphology requires a deep knowledge of the languages involved.To some extent that is true, in that some linguistic knowledge can be useful (detecting mistakes, faster development, problem solving).However, the grammatical aspects under consideration are not so specialised as to require an expert; at least no more than those involved in conventional training. Conclusions The accuracy of the result is strictly dependent on the quality of the map used to train the Morpho Model.Since a good amount of wellformed linguistic data is required to create the map, it is important to handle this correctly.For example, knowing that the customer generally avoids the use of certain verb forms can lead to a reduction in the size of the map, resulting in a simpler task for both the model and the engine.Moreover, the size of the map is a factor that can influence quality and performance.For customers with a small variety of subjects, the map can be reduced based on the words the engine can translate. Further challenges A potential next step for this logic could be to use it in a scenario where both the source and target languages | 3,069,196 | 202670714 | 0 | 16 |
are Slavic.The result could be a greater reduction in vocabulary; however since Slavic languages are quite a homogeneous family, the difference may not be appreciable compared to conventional training. Another interesting field of application might be for languages with non-concatenative morphology, such as Arabic, where words are in-flected with transfixes rather than prefixes or suffixes.The incentive in this case relates not only to the technical challenge, but also to the potential future business opportunities offered by the Middle East and North Africa. | 3,069,197 | 202670714 | 0 | 16 |
Aronia melanocarpa Prevents Alcohol-Induced Chronic Liver Injury via Regulation of Nrf2 Signaling in C57BL/6 Mice Aronia melanocarpa (AM), which is rich in anthocyanins and procyanidins, has been reported to exert antioxidative and anti-inflammatory effects. This study aimed to systematically analyze the components of AM and explore its effects on alcohol-induced chronic liver injury in mice. A component analysis of AM revealed 17 types of fatty acids, 17 types of amino acids, 8 types of minerals, and 3 types of nucleotides. Chronic alcohol-induced liver injury was established in mice via gradient alcohol feeding over a period of 6 months, with test groups orally receiving AM in the last 6 weeks. AM administration yielded potential hepatoprotective effects by alleviating weight gain and changes in organ indexes, decreasing the ratio of alanine aminotransferase/aspartate aminotransferase, reducing lipid peroxidation, enhancing antioxidant activities, decreasing oxidation-related factor levels, and regulating inflammatory cytokine levels. Histological analyses suggest that AM treatment markedly prevented organ damage in alcohol-exposed mice. Furthermore, AM activated nuclear factor erythroid 2-like 2 (Nrf2) by downregulating the expression of Kelch-like ECH-associated protein 1, resulting in elevated downstream antioxidative enzyme levels. AM activated Nrf2 via modulation of the phosphatidylinositol-3-hydroxykinase/protein kinase B signaling pathway. Altogether, AM prevented alcohol-induced liver injury, potentially by suppressing oxidative stress via the Nrf2 signaling pathway. Introduction Alcoholic liver disease (ALD) is a chronic disease worldwide and is associated with increasing mortality rates [1]. In the early stages, ALD typically manifests as steatosis superimposed by an inflammatory infiltrate and progresses to fibrosis or cirrhosis with continued alcohol intake [2]. Excessive | 3,069,198 | 210945092 | 0 | 16 |
drinking can cause alcoholic fatty liver within 2 or 3 weeks and may have further effects on the immune system [3]. Alcohol consumption is also highly correlated with the progression of alcoholic fatty liver [4], causes liver damage, and helps to enhance the production of proinflammatory cytokines and chemokines [5,6], which can enhance the concentration of macrophages and neutrophils for promoting the inflammation response [7]. Long-termed alcohol consumption causes dysfunction within the mitochondrial electron transport chain, resulting in the overgeneration of ROS [8,9]. Furthermore, dyslipidemia resulting from alcohol consumption elicits oxidative and inflammatory responses of varying degrees [10]. Cells possess evolutionarily conserved defensive mechanisms against oxidative stress, including the activation of nuclear factor erythroid 2-like 2 (Nrf2) [11]. The activation of Nrf2-mediated anti-inflammatory pathways is considered an effective way to eliminate excessive ROS [12]. The treatment options for ALD have not changed in the last four decades. Abstinence remains the most effective form of treatment when supported by nutrition and steroids [13]. Limited treatment options are available for patients who are steroid nonresponders or have contraindications to steroid usage (e.g., upper gastrointestinal bleeds, impaired renal functions, and sepsis) [14]. Although corticosteroids (CS) remain a mainstay of treatment for severe alcoholic hepatitis, these drugs are associated with significant risks such as infection, especially in this population of patients [15]. Moreover, the safety of clinical agents still requires rigorous evaluation. Drugs such as metadoxine, an effective drug for hepatic antilipid peroxidation, may induce peripheral neuropathy in patients after long-term use or large dose administration [16]. Thus, safe and | 3,069,199 | 210945092 | 0 | 16 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.