id
stringlengths 24
24
| idx
int64 0
402
| paragraph
stringlengths 106
17.2k
|
|---|---|---|
6617632991aefa6ce1505d28 | 8 | Prior to the initial application, maize plants exhibited similar physiological characteristics, with some germinating earlier than others. The maximum recorded plant height was 61.98 cm, while the minimum was 44.70 cm (Figure , 4 and Table ). Fourteen days after the first application, the percentage increase in average plant height, in |
6617632991aefa6ce1505d28 | 9 | Second application Third application reference to the previous application, followed the order A > B > D > C > E. The highest percentage of average height recorded was 70.98%, while the lowest was 36.12%. When compared to maize D (control), the percentage increase followed the order B > A > C > E, with the highest average height at 44.10% and the lowest at 1.9% |
6617632991aefa6ce1505d28 | 10 | (Figure , 4 and Table ). After the second application, fourteen days later, the percentage increase in average plant height, compared to the previous application, was in the order B > E > C > A > D, with the highest average height at 26.20% and the lowest at 5.80%. When referencing maize D (control), the percentage increase followed the order B > C > A > E, with the highest average height at 71.80% and the lowest at 17.20% which is potassium ratio present in the samples (Table , Figure and). Maize A exhibited stunted growth, yellowing, and leaf burning four days after the second application, which is due to the pH of fertilizer B (table ). Additionally, maize C displayed a dark green leaf colouration, indicating higher chlorophyll content, compared to the rest, following the order C > B > A > E > D, which is due to the total nitrogen content present in all the sample as shown in table . Nitrogen is an essential nutrient for the growth and development of maize plants. It constitutes a vital component of chlorophyll, the pigment responsible for photosynthesis, as well as amino acids, which serve as the building blocks of proteins. 39 After the third application, fourteen days later, the percentage increase in average plant height, compared to the previous application, was in the order A > B > E > D > C. The highest average height recorded was 79.50%, and the lowest was 40.39%. When referencing sample D (control), the percentage increase followed the order A > B > C > E, with the highest average height at 68.11% and the lowest at 19.55% (Table , Figure and). Maize A showed a significantly higher growth percentage compared to maize B, maize C, and maize E when compared to maize D. Its leaves also exhibited an intense dark green coloration, similar to that of maize C. This is attributed to the addition of thiourea as an additional source of nitrogen in the final formulation for maize A as compared to the rabbit manure E. Studies have shown that nitrogen content in typically ranges from 1% to 3% in mature compost. This finding aligns with the results of, who demonstrated that sugarcane waste straw biochar significantly improved okra plant growth and yield parameters when compared to control, farm yard manure, and NPK. It is also consistent with, who reported that the yield of crops using compost from olive mill waste was comparable to that of chemical fertilizers, with no significant difference observed in soil content between the two. |
6617632991aefa6ce1505d28 | 11 | The maize plants (Figure ) exhibited early maturity in maize B, followed by maize A, maize C, maize D, and maize E at the time of harvest as a characteristic difference in the nutrient composition ratios of the samples. The yield was high in Maize A with a total weight of 316.24 g (Table ) which was due to the addition of thiourea, followed by Maize C with a weight of 287.38 g and Maize B 173.17 g respectively. There was minimal difference between maize D and maize E with no signs of grains, indicating limited nutrient in both, possibly both were still in the developmental phase. However, the cob in maize E was slightly larger and taller compared to maize D. This is attributed to slow release of nutrients present in the rabbit manure E, along with insufficient nitrogen and phosphorus content as compared to A, B and C (Table ), as adequate nitrogen supply is crucial for optimal maize growth and yield potential. According to a study by, the application of nitrogen significantly increased the grain yield and biomass production of maize plants, alongside, phosphorus is crucial for root development, seedling growth, and flowering. Maize B, with the highest average plant height of 107.83Β±24 cm had average yield compared to maize A and maize C. Its maturity was faster compared to the other two, likely due to pH as it was still acidic at the time of application and the imbalance of nutrients, particularly nitrogen, although, potassium is crucial for the development of strong stems, roots, and maize cobs. Phosphorus is important for root development, seedling growth, and flowering, while nitrogen is essential for the growth and development of maize plants. It constitutes an important component of chlorophyll, the pigment responsible for photosynthesis, as well as amino acids, which serve as the building blocks of proteins. Consistent with the study by, the application of nitrogen significantly increased the grain yield of the maize plants. Maize A and maize C were nearly identical in size and yield, although maize A outperformed slightly due to higher nitrogen content obtained from thiourea and urea present in the NPK fertilizer. |
6617632991aefa6ce1505d28 | 12 | The residual soil appearance and NPK ratio of the residual soil after harvest is depicted in figure and table 4. To find the percentage of nutrient consumed, we subtract the residual amount (Table ) from the initial application (Table ) for each nutrient and express it as a percentage of the initial application. |
6617632991aefa6ce1505d28 | 13 | For Manure: Nitrogen β 95.04%, Phosphorus β 90.48%, Potassium β 91.19%. These nutrients often require microbial decomposition for release into plant-available forms, resulting in slower uptake and higher residual levels in the soil after harvest. Solubility and Leaching may also contribute to the residual nutrients present after harvest because NPK fertilizers are often highly soluble, which means they dissolve readily in water and are prone to leaching. While this may lead to more efficient nutrient uptake by plants, it can also result in greater nutrient losses through leaching beyond the root zone. On the other hand, nutrients in Fertilizer C and manure may be less soluble or bound within organic matter, making them less susceptible to leaching. Consequently, a portion of these nutrients may remain in the soil after harvest, contributing to higher residual levels. Plants may exhibit varying degrees of efficiency in nutrient uptake depending on factors such as crop type, stage of growth, and environmental conditions. NPK fertilizers, being formulated with balanced nutrient ratios, promote optimal plant growth and development, leading to efficient nutrient uptake and utilization. In contrast, Fertilizer C and manure may provide nutrients in ratios or forms that are not precisely matched to the crop's requirements, potentially resulting in lower uptake efficiency and higher residual levels in the soil. Manure, being an organic fertilizer, relies on microbial decomposition for the release of nutrients into plant-available forms. |
6617632991aefa6ce1505d28 | 14 | This process is typically slower and may continue even after the harvest, leading to sustained nutrient availability in the soil. In contrast, NPK fertilizers provide nutrients in forms that are readily available to plants, bypassing the need for microbial decomposition. Consequently, nutrient release from NPK fertilizers is more rapid, leading to lower residual levels in the soil after harvest. NPK fertilizers tend to leave less residual nutrients in the soil after harvest compared to Fertilizer C and manure due to their balanced nutrient composition, high solubility, efficient plant uptake, and rapid nutrient release. These factors contribute to more effective nutrient utilization by crops and minimize nutrient accumulation in the soil over time although the accumulated nutrient might be highly beneficial to the next crop planted in such soil. Distinct nutrient release patterns were observed from fertilizers B and C, which surpassed the efficacy of NPK fertilizer. Additionally, organic matter persisted as shown in the total carbon content in Table and their appearance in figure , enhancing soil structure, resulting in improved water retention, nutrient holding, and microbial activity-crucial factors for enhancing maize growth. |
6617632991aefa6ce1505d28 | 15 | A successful conversion of plant biomass to organic fertilizer for maize cultivation was achieved, and its performance was compared to inorganic fertilizers, NPK 20:10:5 and rabbit manure. Fertilizer C has a Total nitrogen content of (2.18%), Total phosphorus (1.80 %), Total potassium (3.77 %) and Total Carbon (37.40%) at a pH of 5. Maximum plant height of 171.45 cm for synthesize fertilizer, 134.0 cm for NPK fertilizer, 121.92 cm for rabbit manure and 101.98 cm for control were observed respectively. The most promising treatment was fertilizer C with high a higher percentage of potassium 3.77% compared to NPK fertilizer 5% and with generally higher nutrient concentration compared to rabbit manure. Fertilizer C demonstrated robust plant growth and development even without supplementary inorganic nitrogen in the form of urea. It outperformed all other samples, displaying superior yield compared to rabbit manure. Notably, the addition of 2.55g of thiourea to 19.30g of fertilizer C led to optimal yield, indicating that less than 11% by mass of thiourea addition yields superior results compared to NPK 20:10:5 fertilizers. Fertilizer C also exhibited rapid nutrient release and higher residual organic matter content compared to both NPK and manure. Nutrient consumed for Fertilizer C: N β 91.25%, P β 76.11%, K β 90.64%. for NPK: N β 99.95%, P β 99.94%, P β 99.77%. while for Manure: N β 95.04%, P β 90.48%, K β 91.19% respectively. This underscores the potential for achieving both rapid and slow nutrient release without compromising soil health, as the residual organic matter, a carbon-rich material, decomposes gradually. This underscores that the treatment of agricultural plant residue can significantly influence its suitability as organic fertilizer and soil amendment due to its organic matter content. |
60c73d0ebb8c1ad6163d96a0 | 0 | Due to significant technological advances in the past decades, the body of knowledge on the effects and roles of small molecules in living organisms has grown tremendously . At present, we assume the number of entries across all databases to be in the range of hundreds of millions or billions (10 8 -10 9 ) and a large portion of this data has also accumulated in public databases such as ChEMBL or PubChem BioAssay . Still, these numbers are rather small in comparison to 10 33 , a recently reported estimation of the size of the drug like chemical space . However, it should be noted that numerous studies in the past reported numbers both bigger and smaller depending on the definition used . In addition, considering that only 1-2 measured biological activities per compound are available , the characterization of known compounds also needs to be expanded. |
60c73d0ebb8c1ad6163d96a0 | 1 | For a long time de novo molecular design algorithms for systematic and rational exploration of chemical space and quantitative structure-activity relationship (QSAR) modeling have been considered as tools that could broaden our horizons with less experimental costs and without the need to exhaustively evaluate as many as 10 33 possible drug-like compounds to find the few of interest. The relevance of QSAR modeling and de novo molecular design for drug discovery has been discussed many times , but these approaches can be just as useful in the areas of chemical biology that require new tool compounds and chemical probes that might not be constrained to drug-like molecules only . |
60c73d0ebb8c1ad6163d96a0 | 2 | Thanks to the constant growth of bioactivity databases and widespread utilization of graphical processing units (GPUs) the application of powerful data-driven approaches based on deep neural networks (DNNs) has grown substantially. DNNs found many use cases in molecular virtual screening and de novo compound generation (Figure ) . This rapidly evolving class of algorithms has been influencing modern drug discovery by building more accurate QSAR models , creating better molecular representations , predicting 3D protein structure with impressive accuracy or achieving other promising results in many medicinal and clinical applications . Depending on the architecture, the network is trained either as a bioactivity predictor (e.g. a QSAR model) or as a molecular generator. |
60c73d0ebb8c1ad6163d96a0 | 3 | In the field of de novo drug design, the most attractive feature of DNNs is their ability to probabilistically generate compound structures . DNNs are able to take non-trivial structure-activity patterns into account, thereby increasing the potential for scaffold hopping and the diversity of designed molecules . A large number of generators based on DNNs were developed recently demonstrating the ability of various network architectures to generate compounds of given properties (biological activity included) . |
60c73d0ebb8c1ad6163d96a0 | 4 | Even though deep learning has been dominating de novo drug design in the recent years, it should be noted that the field also has a long history of evolutionary heuristic methods such as genetic algorithms on the forefront . These traditional methods are still being investigated and developed and it is yet to be established how they compare to the novel approaches based on deep learning . Due to the simpler nature of these traditional approaches non-obvious relationships can be easily missed, which may affect the quality of the suggested chemical structures. However, simplicity can also be an advantage since interpretation of simpler methods is easier. This is especially problematic for deep learning models that can have more than thousands of parameters . Moreover, a simpler method requires less training data without sacrificing chemical space coverage . |
60c73d0ebb8c1ad6163d96a0 | 5 | One of the open questions for both traditional and deep learning molecular generators is also how they should be benchmarked, compared and interpreted . Therefore, benchmarking studies of de novo drug design approaches are also the subject of ongoing research and much needed to ensure that these methods have conclusive real impact on new ligand discovery . However, the ultimate test of a de novo drug design method should always be prospective application in real projects with experimental validation of the generated molecules. |
60c73d0ebb8c1ad6163d96a0 | 6 | Although de novo molecular design algorithms have been in development for multiple decades and experimentally validated active compounds have been proposed , these success stories are still far away from the envisaged performance of the 'robot scientist' . Successful development of a completely automated and sufficiently accurate process has been elusive and hindered mostly by the computational expense and poor synthetic availability of the generated compounds . Despite increasing efforts to automate the scientific process of decision making , human insight and manual labor are still necessary to further refine the compounds generated by de novo molecular design algorithms. |
60c73d0ebb8c1ad6163d96a0 | 7 | Though many in silico compound generation and optimization tools are available for free , it is still an exception that these approaches are routinely used. The vast majority of methods described in the literature serve only as a proof of concept. Hence, they lack a proper graphical user interface (GUI) through which non-experts could easily access the algorithms and analyze their inputs and outputs in a convenient way. Even if such a GUI exists, it is often simplistic and intended to be used only with one particular method . Lack of easy to explain and auditable information systems is a factor leading to some level of disconnection between medicinal and computational chemists , which can hinder tighter collaboration that can stand in the way of effective utilization of many promising de novo drug design methods. Many molecular generators would also benefit from a comprehensive and easy to use application programming interface (API) that would enable easier integration with existing computational infrastructures. Recently a tool called Flame was presented that offers many of the aforementioned features in the field of predictive QSAR modeling , but while there are closed-source solutions like BRADSHAW or Chemistry42 to the best of our knowledge there is no such solution in the realm of open source software for de novo drug design. However, there has been effort to develop interactive databases of generated structures as evidenced by the most recent example, cheML.io . |
60c73d0ebb8c1ad6163d96a0 | 8 | In this work we present the development of GenUI, a software framework that provides a general-purpose GUI for molecular generators and enables easy integration of such algorithms with existing drug discovery pipelines as well. The GenUI framework integrates solutions for import, generation, storage and retrieval of compounds, visualization of the created molecular data sets and basic utilities for QSAR modeling. All features can be easily accessed through the web-based GUI or REST API to ensure that both human users and automated processes can interact with the application easily. Integration of new molecular generators and other features is facilitated by a Python API and GUI customization is possible via custom components implemented with the React.js JavaScript library. To demonstrate the features of the GenUI framework, our recently published molecular generator DrugEx was integrated within the GenUI ecosystem. The source code of the GenUI platform is distributed under the MIT open-source license [75-77] and several Docker images are also available online for quick deployment . |
60c73d0ebb8c1ad6163d96a0 | 9 | User interaction with GenUI happens through the frontend web client which issues REST API calls to the backend, which comprises five services (Figure ). However, advanced users may also implement clients and automated processes that use the REST API directly. so that any automated process can also interact with GenUI. The backend application comprises five REST API services each of which has access to the data storage and task queue subsystems. The services can issue computationally intensive and long-running asynchronous tasks to backend workers to ensure sufficient responsiveness and scalability. In the current implementation, tasks can be submitted to two queues: (1) the default CPU queue, which handles all tasks by default, or (2) the GPU queue, intended for tasks that can be accelerated by the use of GPUs. |
60c73d0ebb8c1ad6163d96a0 | 10 | The GUI is implemented as a JavaScript application built on top of the React.js web framework. The majority of graphical components is provided by the Vibe Dashboard opensource project , but the original collection of Vibe components was considerably expanded with custom components to fetch, send, and display data exchanged with the GenUI backend REST API. In addition, frameworks Plotly.js , Charts.js and ChemSpace.js are used to provide helpful interactive figures. |
60c73d0ebb8c1ad6163d96a0 | 11 | The GUI reflects the structure of the GenUI backend services (Figure and Figure ). Each backend service (Projects, Compounds, QSAR, Generators, and Maps) is represented as a separate item in the navigation menu on the left side of the interface (Figure ). Upon clicking a menu item, the corresponding page opens rendering a grid of cards (Figure ) that represent the objects corresponding to the selected backend service. Various actions related to the particular service can be performed from the action menu in the top right of the interface (Figure ). |
60c73d0ebb8c1ad6163d96a0 | 12 | The "Projects" interface serves as a simple way to organize user workflows. For example, a project can encapsulate a workflow for the generation of novel ligands for one protein target (Figure ). Each project contains imported compounds, QSAR models, molecular generators and chemical space maps. The number of projects per user is not limited and they can be deleted or created as needed. Receptor" project is already open so the menu on the left can be used to access its data. The GUI consists of three main parts: a) navigation menu, b) card grid and c) action menu. The navigation menu is used to browse data associated with various GenUI services ("Projects" in this case). If a link is clicked in the navigation menu, the data of the selected service is displayed as a grid of interactive cards. Each card allows the users to manage particular data items (a project in this case). The action menu in the top right is also updated depending on the service selected in the navigation menu and performs actions not related to a particular data item. In this case, the action menu was used to bring up the project creation form on the bottom left of the card grid. |
60c73d0ebb8c1ad6163d96a0 | 13 | Compounds can be generated by generators, but also imported from SDF files, CSV files or obtained directly from the ChEMBL database . New import filters can be easily added by extending the Python backend and customizing the components of the React API accordingly (see Python API and JavaScript API). For each compound in the compound set the interface can display its 2D representation (Figure ), molecular identifiers (i.e. SMILES, InChI, and InChIKey), reported and predicted activities (Figure ) and physicochemical properties (i.e. molecular weight, number of heavy atoms, number of aromatic rings, hydrogen bond donors, hydrogen bond acceptors, logP and topological polar surface area). |
60c73d0ebb8c1ad6163d96a0 | 14 | Figure A screenshot showing part of the "Compounds" GUI. In this page, users can import data sets from various sources. A card representing an already imported data set from the ChEMBL database is shown. The position and size of each displayed card can be modified by either dragging the card (reposition) or adjusting the bottom right corner (size change). The card shown is currently expanded over two rows of the card grid (Figure ) in order to accommodate the displayed data better. The "Activities" tab in the compound overview shows summary of the biological activity data associated with the compound. The activities are grouped by type and aside from experimentally determined activities the interface also displays activity predictions of available QSAR models. For example, in the view shown the "Active Probability" activity type is used to denote the output probability from a classification QSAR model. Each activity value also contains information about its origin (the "Source" column) so that it can be tracked back to its source. |
60c73d0ebb8c1ad6163d96a0 | 15 | The "Performance" tab lists various performance measures of the QSAR model obtained by cross-validation or on an independent hold out test set (Figure ). The validation procedure can be adjusted by the user during model creation (Figure ). Making predictions with the model is possible under the "Predictions" tab. Each QSAR model can be used to make predictions for any compound set listed on the "Compounds" page and the calculated predictions will then become visible in that interface as well (Figure ). New QSAR models are submitted for training with a creation card (Figure ) that helps users choose model hyperparameters and a suitable training strategy (i.e. the characteristics of the independent hold out validation set, the number of cross-validation folds or the choice of validation metrics). The "Info" tab of a trained model contains important metadata as well as a hyperlink to export the model and save it as a reusable Python object. This import/export feature enables users to archive and share their work, enhancing the reusability and reproducibility of the developed models . The "Performance" tab can be used to observe model performance data according to the chosen validation scheme (Figure ). This information is different depending on the chosen model type (regression vs. classification, Figure vs. Figure ) and the parameters used (i.e. the choice of validation metrics). |
60c73d0ebb8c1ad6163d96a0 | 16 | Under the "Generators" menu item, the users find a list of individual generators implemented in the GenUI framework (Figure ). Currently, only the DrugEx generator is available, but other generators can be added easily by extending the Python backend and customizing the existing React components. In fact, the GUI for DrugEx is based on the same React components as the "QSAR Models" view. Like QSAR models, DrugEx networks can also be serialized and saved as files. For example, a cheminformatics researcher can build a DrugEx model outside of the GenUI ecosystem (i.e. using a script published with the original paper ) and provide the created model files to another researcher who can import and use the model from the GenUI web-based GUI. |
60c73d0ebb8c1ad6163d96a0 | 17 | Interactive visualization of chemical space is available under the "Maps" menu item. The menu separates the creation of the chemical space visualization, the "Creator" page (Figure ), and its exploration, the "Explorer" page (Figure ). Figure The "Creator" interface of GenUI "Maps" page. On the left a form to create a new t-SNE mapping of two sets of compounds using Morgan fingerprints is shown while information about an existing map can be seen on the right. |
60c73d0ebb8c1ad6163d96a0 | 18 | It is possible to color points by biological activities, physicochemical properties and other data associated with the compounds. The same can also be done with the size of the points. The points drawn in the map are interactive and hovering over a point shows a box with information about the compound inside and on the right side of the map. Groups of points can also be selected by drawing a rectangle over them in which case a list of selected compounds is shown in the "Selection List" tab (Figure ) and their bioactivity data is summarized under the "Selection Activities" tab (Figure ). |
60c73d0ebb8c1ad6163d96a0 | 19 | The "Creator" page is implemented as a grid of cards each of which represents an embedding of chemical compounds in 2D space (Figure ). Implicitly, the GenUI platform enables t-SNE embedding (provided by openTSNE ). However, new projection methods can be easily added to the backend through the GenUI Python API with no need to modify the GUI (see Python API) . |
60c73d0ebb8c1ad6163d96a0 | 20 | The purpose of the "Explorer" page is to interactively visualize chemical space embedding prepared in the "Creator" (Figure ). In the created visualization the users can explore compound bioactivities, physicochemical properties, and other measurements for various representations and parts of chemical space. Thanks to ChemSpace.js up to 5 dimensions can be shown in the map at the same time: X and Y coordinates, point color, point size and point shape. The map can be zoomed in by drawing a rectangle over a group of points. Such points form a selection and their detailed information is then displayed under the "Selected List" (Figure ) and "Selected Activities" tabs (Figure ). Figure View of the "Selected List" tab of the "Explorer" page. The tab shows the selected molecules in the map as a list which is the same as the one used in the "Compounds" view (Figure ). For easier navigation, the compounds are also grouped by the compound set they belong to and the view for each set can be accessed by switching tabs above the displayed list (only one compound set, CHEMBL251, is present in this case). Therefore, the frontend GUI comprises a large library of over 50 React components that are encapsulated in a standalone package (Figure ). The package is organized into subpackages that follow the structure and hierarchy of design elements in the GenUI interface. |
60c73d0ebb8c1ad6163d96a0 | 21 | In the following sections, we use the two most important groups of the React API components as case studies to illustrate how the frontend GUI can be extended. The presented components are "Model Components", used to add new trainable models, and the "REST API Components", used to fetch and send data between the frontend and the GenUI REST API services. |
60c73d0ebb8c1ad6163d96a0 | 22 | Figure Schematic depiction of the GenUI React library which contains customized styles, utility functions and the components used in the GenUI web client. The "components" subpackage organizes the components into groups related to the structure of the GenUI interface. For example, components filed under the "models" subpackage are used in the creation of the "QSAR Models" (Figure ), "DrugEx" (Figure ) and "Maps" (Figure ) interfaces while components under the "compounds" subpackage are used to implement the "Compounds" view (Figure ). General purpose components (i.e. the card grid or the card tab widget) are in the root of the "components" subpackage. |
60c73d0ebb8c1ad6163d96a0 | 23 | Much of the functionality of the GenUI platform is based on trained models. The "QSAR Models", "DrugEx" and "Maps" pages all borrow from the same library of reusable GenUI React components (Figure ). At the core of the "models" component library (Figure ) is the ModelsPage component (Figure ). ModelsPage manages the layout and data displayed in model cards. When the users select to build a new model, the ModelsPage component is also responsible to show a card with the model creation form. The information that the ModelsPage displays can be customized through various React properties (Figure ) that represent either data (data properties) or other components (component properties). Such an encapsulation approach and top-down data flow is one of the main strengths of the React framework. This design is very robust since it fosters appropriate separation of concerns by their encapsulation inside more and more specialized components. This makes the code easy to reuse and maintain. ModelsPage). This creates a hierarchy of reusable components that can be easily assembled and configured to accommodate the different needs of each model view in a standardized and consistent manner. |
60c73d0ebb8c1ad6163d96a0 | 24 | Because the GUI often needs to fetch data from the backend server, several React components were defined for that purpose. In order to use them, one just needs to provide the required REST API URLs as React component properties. For example, the ComponentWithResources component configured with the '/maps/algorithms/' URL will get all available embedding methods as JSON and converts the result to a JavaScript object. Many components can also periodically update the fetched data, which is useful for tracking information in real time. For paginated data there is also the ApiResourcePaginator component that only fetches a new page if a given event is fired (i.e. user presses a button). This makes it convenient to create GUIs for larger data sets. In addition, user credentials are also handled automatically. |
60c73d0ebb8c1ad6163d96a0 | 25 | Many more specialized components are also available to fetch specific information. For example, the TaskAwareComponent tracks URLs associated with background asynchronous tasks and it regularly passes information about completed, running, or failed tasks to its child components. However, other specialized components exist that automatically fetch and format pictures of molecules, bioactivities, physicochemical properties or create, update and delete objects in the UI and the server . |
60c73d0ebb8c1ad6163d96a0 | 26 | The backend services are the core of the GenUI platform and the GenUI Python API provides a convenient way to write backend extensions (i.e. new molecular generators, compound import filters, machine learning algorithms for QSAR modeling, and dimensionality reduction methods for chemical space maps). All five backend services (Figure ) are implemented with the Django web framework and Django REST Framework . For data storage, a freely available Docker image developed by Informatics Matters Ltd. is used. The Docker image contains an instance of the PostgreSQL database system with integrated database cartridge from the RDKit cheminformatics framework . The integration of RDKit with the Django web framework is handled with the Django RDKit library . All compounds imported in the database are automatically standardized with the current version of the ChEMBL structure curation pipeline . |
60c73d0ebb8c1ad6163d96a0 | 27 | The GenUI backend codebase is divided into multiple Python packages that each encapsulate a part of the GenUI Django project (Figure ). Any package that resides in the root directory is referred to as the root package. Root packages facilitate many of the REST API endpoints (Figure ), but they also contain reusable classes that are intended to be extended by extensions (see Generic Views and Viewsets, for example). In the following sections, some important features of the backend Python API are briefly highlighted. However, a much more detailed description with code examples is available on the documentation page of the project . |
60c73d0ebb8c1ad6163d96a0 | 28 | Just like in the case of the GenUI React API, modularity and extensibility were also the main concerns during the design of the GenUI backend services. Each of the aforementioned root packages contains a Python package called extensions (Figure ). The extensions package can contain any number of Django applications or Python modules, which ensures that the extending components of the GenUI framework are well-organized and loosely coupled. |
60c73d0ebb8c1ad6163d96a0 | 29 | Provided that GenUI extensions are structured a certain way they can take advantage of automatic configuration and integration (see Automatic Code Discovery). Before the Django project is deployed, GenUI applications and extensions are detected and configured with the genuisetup command, which makes sure that the associated REST API endpoints are exposed under the correct URLs. The genuisetup command is executed with the manage.py script (a utility script provided by the Django library). |
60c73d0ebb8c1ad6163d96a0 | 30 | The root packages of the GenUI backend library define many abstract and generic base classes to implement and reuse in extensions. These classes either implement the REST API or define code to be run on the worker nodes inside Celery tasks. Automatic code discovery uses several introspection functions and methods to find the derived classes of the base classes found in the root packages. By default, this is done when the genuisetup command is executed (see Extensions). |
60c73d0ebb8c1ad6163d96a0 | 31 | For example, if the derived class defines a new machine learning algorithm to be used in QSAR modelling, automatic code discovery utilities make sure that the new algorithm appears as a choice in the QSAR modelling REST API and that proper parameter values are collected via the endpoint to create the model. Moreover, all changes also get automatically propagated to the web-based GUI because it uses the REST API to obtain algorithm choices for the model creation form. Thus, no JavaScript code has to be written to integrate a new machine learning algorithm. These mechanisms are also used when adding molecular generators, dimensionality reduction methods, or molecular descriptors. |
60c73d0ebb8c1ad6163d96a0 | 32 | When developing REST API services with the Django REST Framework, a common practice is using generic views and sets of views (called viewsets). In Django applications, views are functions or classes that handle incoming HTTP requests. Viewsets are classes defined by the Django REST Framework that bring functionality of several views (such as creation, update or deletion of objects) into one single class. Generic views and viewsets are then classes that usually do not stand on their own, but are designed to be further extended and customized. |
60c73d0ebb8c1ad6163d96a0 | 33 | The GenUI Python library embraces this philosophy and many REST API endpoints are encapsulated in generic views or viewsets. This ensures that the functionality can be reused and that no code needs to be written twice, as stated by the well-known DRY ("Don't Repeat Yourself") principle . An example of such a generic approach is the ModelViewSet class that handles the endpoints for retrieval and training of machine learning models. This viewset is used by the qsar and maps applications, but also by the DrugEx extension. All these applications depend on some form of a machine learning model so they can take advantage of this interface, which automatically checks the validity of user inputs and sends model training jobs to the task queue. |
60c73d0ebb8c1ad6163d96a0 | 34 | Many of the GenUI backend services take advantage of asynchronous tasks which are functions executed in the background without blocking the main application. Moreover, tasks do not even have to be executed on the same machine as the caller of the task, which allows for a great deal of flexibility and scalability (see Deployment). |
60c73d0ebb8c1ad6163d96a0 | 35 | The Celery task queue makes creating asynchronous tasks as easy as defining a Python function [100]. In addition, some GenUI views already define their own tasks and no explicit task definition is needed in the derived views of the extensions. For example, the compounds root package defines a generic viewset that can be used to create and manage compound sets. The import and creation of compounds belonging to a new compound set is handled by implementing a separate initializer class, which is passed to the appropriate generic viewset class . The initialization of a compound set can take a long time or may fail and, thus, should be executed asynchronously. Therefore, the viewset of the compounds application automatically executes the methods of the initializer class asynchronously with the help of an available Celery worker. |
60c73d0ebb8c1ad6163d96a0 | 36 | Since the GenUI platform consists of several components with many dependencies and spans multiple programming languages, it can be tedious to set up the whole project on a new system. The tools to build these images are freely available . Therefore, developers can create images for extended versions of the GenUI that fit the needs of their organizations. In addition, the separation of the main application (genui-main) from workers also allows distributed deployment over multiple machines, which opens up the possibility to create a scalable architecture that can quickly accommodate teams of varying sizes. |
60c73d0ebb8c1ad6163d96a0 | 37 | Although the GenUI framework already implements much of the functionality needed to successfully integrate most molecular generators, there are still many aspects of the framework that can be improved. For instance, it would be beneficial if more sources of Even though it is hard to determine the requirements of every project where molecular generators might be applied, many of the aforementioned features and improvements can be readily implemented with the GenUI React components (see JavaScript API) and the Python API (see Python API). In fact, the already presented extensions and the DrugEx interface are useful case studies that can be used as templates for integration of many other cheminformatics tools and de novo molecular generators. Therefore, we see GenUI as a flexible and scalable framework that can be used by organizations to quickly integrate tools and data the way it suits their needs the most. However, we would also like GenUI to become a new useful way to share the progress in the development of novel de novo drug design methods and other cheminformatics approaches in the public domain. |
651713f1ade1178b245e6e88 | 0 | The rise of intermittent renewable energy generation and vehicle electrification has created exponential growth in lithium-ion battery (LIB) production beyond consumer electronics. By 2030, the electric vehicle (EV) sector is projected to dominate LIB growth, accounting for 82% of an estimated 2.4 TWh yr -1 of total global LIB production (Fig. , Supplementary Information). |
651713f1ade1178b245e6e88 | 1 | However, the limited supply of critical materials (e.g., Li, Ni, Co, and Cu 1 ) needed for prominent LIB chemistries has exacerbated environmental, economic, national security, and human rights concerns . Critical LIB materials are projected to reach major global supply-demand balance deficits before 2030 (Fig. ). Further, both mining of LIB materials and improper disposal of end-of-life LIBs can damage natural and human ecosystems, cause occupational hazards during handling, and result in monetary losses . |
651713f1ade1178b245e6e88 | 2 | Recycling critical materials in end-of-life LIBs can help alleviate growing environmental concerns and is essential for the long-term sustainability of electrified transportation. While recycled materials may not contribute substantially to global LIB demand for decades, the establishment of domestic circular supply chains is iterative, requiring multiple learning curves as the dominant supply of end-of-life LIB chemistries and form factors evolve and as supply grows. |
651713f1ade1178b245e6e88 | 3 | Factors central to the success of recycling include the ease of collecting products, the cost of recycling processes, and the economic value of recovered materials. The average embodied economic values of representative LIBs between 2018-2021 are shown in Fig. (complete references are listed in Supplementary Information). In LIBs, between 2018-2021, Li, Ni, and Co comprise the highest embodied economic value, and Al and Cu account for a significant weight percentage of EV battery packs (approximately 25%) . Despite an embodied economic value that is 2-10 times higher compared to the lead in lead-acid batteries, LIBs are only recycled 2-47% globally , compared to 99% for lead-acid batteries in the U.S. Regardless, the untapped potential of LIB recycling constitutes a significant economic and environmental opportunity that requires evaluation across several application scales, from numerous small-scale consumer electronic LIBs (e.g., 10-100 Wh) to fewer large-scale transportation and stationary storage LIB packs (e.g., 10-100 kWh) . In addition, the preferred chemistries by automakers have evolved to hedge potential critical mineral shortages and react to market shifts, such as the near tripling of lithium carbonate prices in early 2022. Existing LIB variation and supply chain complexity highlight the need for a methodical and comparative life cycle assessment (LCA) between circular (i.e., recycling used batteries) and conventional supply chains, which is also necessary for future recycling of the evolving portfolio of battery chemistries. Representative LIBs are from consumer electronics using lithium cobalt oxide (LCO), and electric vehicle battery packs including lithium nickel manganese cobalt oxide (NMC111 and NMC811), lithium nickel cobalt aluminum oxide (NCA), lithium manganese oxide (LMO), and lithium iron phosphate (LFP). Data are based on market values adjusted for inflation between January 2018 and December 2021 (complete references are listed in Fig. in Supplementary Information), and the uncertainty denotes a 90% confidence interval, which may overlap with the data point in some instances, obscuring their view. The blue shaded area in the upper panel represents the average commodity values of commonly recycled products: glass, paper, plastic, and metal cans (more details are provided in Fig. ). b, Cradle-to-gate steps of manufacturing battery-grade LIB materials (i.e., salts) from conventional and circular supply chains, both of which include three steps: extraction, transport, and refinement. Extraction and transport are considered upstream steps relative to gate-to-gate refinement, which is indicated by the red shaded area between "input" and "output" gates. Cradle-to-gate analysis considers the refinement and upstream processes together. |
651713f1ade1178b245e6e88 | 4 | Despite significant progress, current understanding of the relative environmental impacts of recycling LIBs is still incomplete. The most significant environmental differences between LIB production from circular and conventionally mined cathode material lie early in supply chains, comprised of extraction, transport, and refinement steps (together "cradle-to-gate," Fig. ). While several previous studies have investigated cradle-to-gate environmental impacts, gate-to-gate analyses of circular refinement processes are inconsistent, reporting environmental impacts that differ by >30% , and are not completely based on industrial-scale LIB recycling operations. The gate-to-gate refinement processes utilized at established and emerging circular refinement facilities may include mechanical separation (Me), pyrometallurgy (Py), and hydrometallurgy (Hy) . Specifically, Me physically dismantles LIBs into constituent components, Py leverages elevated temperature to facilitate thermally-driven material transformations, and Hy separates materials in the aqueous phase via leaching, precipitation, and solvent extraction processes. |
651713f1ade1178b245e6e88 | 5 | Variations in environmental impacts arise from the specific operational choices at refinement facilities that utilize different processing pathways and from the methods to evaluate them. There is a critical need for transparency and detailed insights into the environmental impacts (e.g., energy consumption, greenhouse gas emission, and water consumption) of LIB refinement pathways and all cradle-to-gate supply chain steps. Previous efforts have worked towards addressing this need , and this study builds on the comparative methodology of a recent step-by-step study to provide higher resolution and more actionable primary data, insights, and recommendations. |
651713f1ade1178b245e6e88 | 6 | In this study we quantify the cradle-to-gate environmental impacts of battery-grade cathode material salts manufactured in conventional and circular supply chains across three major steps: material extraction, transport, and refinement (Fig. ). First, we quantify and compare the refinement of mined concentrate from natural deposits into battery-grade materials in conventional supply chains with production of these materials by Redwood Materials (a recycling company in Nevada, U.S.). Two LIB feedstocks are explored: non-energized LIB production scrap from manufacturing facilities and energized end-of-life LIBs collected from consumers. Industrial-scale operational data provided by Redwood Materials are analyzed and compared to conventional LIB supply chain values based on Argonne National Laboratory's Greenhouse Gases, Regulated Emissions, and Energy use in Technologies (GREET 2021) model . Second, influences of the product formats in the refinement pathways on environmental impacts are examined. For both conventional and circular refinement, impacts of producing mixed Ni-Co compounds and discrete salts are analyzed. Third, we assess the environmental impacts of upstream processes before gateto-gate refinement based on modeling. The upstream assessment includes the extraction of LIB material from conventional (i.e., mined ore) or circular (i.e., collected batteries) sources and the transport of extracted material to relevant refinement facilities for production of battery-grade cathode materials as Li, Co and Ni sulfate or carbonate salts. To the best of our knowledge, this study is the first life cycle assessment with primary industrial-scale circular refinement data that includes stepwise, cradle-to-gate comparison of conventional and circular LIB supply chains. With the methodologies and results reported herein, researchers can prioritize major opportunities to improve process efficiencies, practitioners can benchmark their environmental impacts, and policymakers can incentivize best environmental practices in LIB supply chain management. |
651713f1ade1178b245e6e88 | 7 | In LIB supply chains, the refinement step converts the collected feedstocks into battery-grade salts for further manufacturing (Fig. ). In both conventional and circular supply chains, the refinement pathways vary significantly depending on multiple factors. Five refinement pathways are compared in this study (Fig. ). While conventional refinement starts with mined ores/brines (1 and 2), circular refinement starts with either end-of-life batteries (1 and 2) or battery scrap (5). |
651713f1ade1178b245e6e88 | 8 | Ni and Co in refinement products for subsequent manufacturing can be discrete salts (1 and 3) or mixed compounds (2, 4, and 5). Refining lithium-ion batteries into battery-grade materials exhibits lower environmental impacts than production from mined natural materials. The upstream steps of material extraction and transport are considered in later sections. Environmental impacts including energy consumption, greenhouse gas emissions (CO2-equivalents, CO2-eq; additional criteria air pollutants are detailed in Table ) and water consumption of refinement pathways in conventional and circular LIB supply chains are compared in Fig. for the gate-to-gate production of battery-grade cathode materials. State-of-the-art conventional pathways generating discrete salts (Method (1) in Fig. ) are analyzed here. One kg of lithium-nickel-cobalt-aluminum-oxide cathode-equivalent material (NCA-eq) is employed as a functional unit throughout this study for supply chain comparison, accounting for the elemental requirements to produce stoichiometric LiNi0.80Co0.15Al0.05O2. NCA chemistry is selected for the functional unit because it comprises the second-largest category of EV battery chemistries following NMC batteries , and is projected to utilize less Co compared to NMC . Excluding the environmental impacts of material extraction and transport steps, the gate-to-gate production of one kg NCA-eq battery-grade material from conventional mined natural materials consumes 193.9 MJ and 77.3 L H2O while emitting 14.5 kg CO2-eq (Fig. ). The values of energy consumption and greenhouse gas emissions are comparable with previous studies based on GREET datasets (Fig. ). Refinement of mined material concentrate into battery-grade Ni material dominates NCA environmental impacts, representing >57% of total values. note that Al is presented on the top of the stacked bars but its contributions are too small to be seen; however specific environmental impacts of each element contributor are detailed in Table . |
651713f1ade1178b245e6e88 | 9 | Multi-step circular refinement pathways include mechanical processing (Me, grey), reductive calcination (RC, red), and hydrometallurgy (Hy, blue). RC is an additional processing step for The environmental impacts of two circular refinement pathways are presented in each graph in Fig. for mixed-stream LIB feedstocks processed at Redwood Materials: non-energized production scrap from LIB production facilities (recycled scrap) and energized, end-of-life LIBs collected from consumers (recycled battery). Using a limiting-reagent approach of output products to produce one kg NCA-eq material, energy requirements for processing recycled scrap and recycled battery streams are 22.0 MJ/kg and 44.4 MJ/kg NCA-eq materials, significantly lower than conventional refinement by 88.7% and 77.1%, respectively (Fig. ). Relatedly, 2.9 and 6.9 kg CO2-eq/kg NCA-eq materials are generated from scrap and battery streams, respectively, a substantial reduction in CO2-eq emissions by 80.0% and 59.1% (Fig. ). Water consumption is also lower by 88.4% for scrap and 74.1% for battery streams relative to the conventional scenario, resulting from the consumption of 9.0 and 20.0 L H2O/kg NCA-eq materials, respectively (Fig. ). Note that while the elemental stoichiometry is identical, the output battery-grade materials vary slightly between conventional (Li2CO3, NiSO4, CoSO4) and circular (Li2SO4, (Ni, Co)SO4) supply chains (detailed in Methods). Converting the final lithium product to Li2CO3 does not substantially change the environmental impacts of the circular supply chains (Supplementary Note 3, Fig. ), and impacts of producing discrete or mixed products are examined in the following section. |
651713f1ade1178b245e6e88 | 10 | To produce battery-grade cathode materials, Redwood Materials uses a combination of reductive calcination (RC), mechanical (Me), and hydrometallurgical (Hy) LIB refinement processes (pathways detailed in Fig. ). The RC process converts energized battery feedstock under certain conditions that leverage heat from exothermic processes and inhibit graphite combustion. This process does not use direct fossil fuel inputs onsite and facilitates subsequent hydrometallurgical refinement into battery-grade materials. Because RC is not required for nonenergized LIB production scrap materials, the two feedstock streams (recycled scrap and recycled batteries) are analyzed separately. Energy consumption and CO2-eq emissions of representative existing recycling pathways from the literature, including pyrometallurgy (Py*), hydrometallurgy (Hy*), and direct recycling (Direct*), are also presented in Fig. for comparison. In general, the RC+Me+Hy pathway at Redwood exhibits comparable energy consumption and CO2-eq emissions with Hy and Direct literature values , and substantially lower environmental impacts than Py*. |
651713f1ade1178b245e6e88 | 11 | Note that traditional pyrometallurgy and Redwood Material's reductive calcination can process energized batteries of varying states of charge, health, and formats with minimal modification, whereas traditional hydrometallurgy may need to discharge energized batteries in salt bath or cryogenically remove electrolyte for safe mechanical processing. While this analysis is focused on Redwood Materials refinement pathways, the methodology can be used to evaluate additional refinement pathways (e.g., hydrometallurgy in Fig. ), or others that use different material feedstocks, refinement processes, and energy supplies. |
651713f1ade1178b245e6e88 | 12 | Among the few studies that directly compare environmental impacts of circular and conventional NCA refinement using industrial-scale operational data, 35% lower greenhouse gas emissions (Fig. ) are reported for Me+Hy circular refinement compared with the current study . However, direct comparison can be inexact due to varying underlying assumptions and data sources. For example, Argonne National Laboratory's GREET and EverBatt models leverage a combination of technology descriptions from patent applications (the most recent from 2007), literature data on process flow consumptions, industry site visits and surveys, expert advice solicitation, and stated assumptions to form complete pathways. Further, Ciez and Whitacre quantified environmental impacts using output products represented as "metal offsets" for pyrometallurgy or with metals in solution for hydrometallurgy (Note 3 in Supplementary Information), rather than cathode salts in this study. In addition, the previous studies included a portion of recycled metal materials in its conventional supply chain analysis, whereas this work references only mined natural deposits in conventional supply chains to fully deconvolute the environmental impacts . The different conclusions highlight divergent life cycle assessment approaches, processing conditions, and the utility of primary industrial data access over modeling processes from literature sources. |
651713f1ade1178b245e6e88 | 13 | Ni and Co are key elements for battery manufacturing, and can be traded in the format of mixed metal salts or discrete salt products between battery refiners and battery manufacturers . To examine the influences of the refinement product formats, the environmental impacts of refinement to mixed salt are compared to the refinement to discrete sulfate salts, NiSO4 and CoSO4 (Fig. ). |
651713f1ade1178b245e6e88 | 14 | The GREET model is employed to analyze different conventional mining pathways generating different product formats (detailed in Methods). In conventional mining, refining Ni-Co ores to mixed hydroxide precipitate, (Ni,Co)(OH)2 (Method (2) in Fig. ), elevates energy consumption and CO2-eq emissions by 77.% and 89.4%, respectively, over the discrete salts-based pathway (Fig. and, left panels). While the discrete products NiSO4 and CoSO4 are produced from the mixed hydroxide precipitates through additional post-treatment, the very low composition of Co (3.6%) in the latter limits the NCA stoichiometry, thus increasing the total energy cost to generate 1 kg NCA-equivalent materials. On the other hand, water consumption of refining mixed hydroxides is slightly lower (-6.6%) than that in producing discrete salts. |
651713f1ade1178b245e6e88 | 15 | Circular pathways refining batteries to different products are analyzed using the Redwood data by the RC+Me+Hy process and the modeling of a representative battery recycling method combining mechanical and hydrometallurgy (Me+Hy) refinement (Method (3) in Fig. ). The Redwood process refines recycled batteries to mixed metal sulfate, (Ni,Co)SO4, whereas the representative Me+Hy produces discrete NiSO4 and CoSO4 as the products. The RC pathway (RC+Me+Hy) exhibits lower energy consumption (-72.3%), CO2-eq emissions (-39.5%), and water consumption (-12%) relative to the Me+Hy pathway (Fig. ), because it avoids additional treatment separating (Ni,Co)SO4 to discrete salts. Overall, our results indicate that refining batteries to mixed metal salts instead of discrete salts can substantially save environmental impacts while still satisfying the needs of circular LIB supply chains. Our findings also provide important insights to optimizing plant-scale battery refining operations. In the following sections, mixed saltbased pathways are analyzed for refinement. ). Numbers in parentheses labelled on the top of stacked bars denote the refinement methods summarized in Fig. . |
651713f1ade1178b245e6e88 | 16 | Electricity consumption dominates the environmental impacts of lithium-ion battery circular refinement. The relative environmental impacts of input consumables (e.g., energy, water, commodity chemicals) in the gate-to-gate refinement processes are disaggregated in Fig. (additional criteria air pollutants in Tables S2-S3, Figs. ). Note that the embodied environmental impacts of electricity consumption in Fig. are based on the Nevada Power Company (NEVP) at the Redwood Materials location. Electricity consumption is a principal factor dominating the environmental impacts. For both LIB feedstock pathways (Methods (4) and (5) in Fig. ), electricity accounts for 70.3-91.0% of the total energy consumption, 71.8-79.1% of the total CO2-eq emissions, and 54.3-63.6% of water consumption (Fig. ). For both feedstocks, Hy processes comprise the majority of environmental impacts, contributing more than 87.3% to energy consumption, 86.3% to CO2-eq emission, and 88.8% to water consumption. Notably, the additional RC step required for processing energized batteries only marginally contributes to CO2eq emissions (7.4% of total). Unlike conventional pyrometallurgical processes that require external energy sources , the RC process is primarily autothermic because it leverages process heat released from exothermic reactions of the LIB materials . In addition to electricity consumption, chemical reagents used in circular refinement processes also contribute to embodied environmental impacts. Alkali reagents used to precipitate metals contribute between 19.0-21.3% of environmental impacts (largest relative contribution to water consumption). H2O2 is used to reduce high-oxidation state metal compounds for hydrometallurgical leaching of scrap material, and accounts for 11.3-20.1% of environmental impacts (largest relative contribution to energy consumption). |
651713f1ade1178b245e6e88 | 17 | . c, Tradeoff relationship between embodied water consumption and CO2-eq emission by different power sources, including electricity grids in different locations (β), purely power sources (β‘), and Nevada renewable energy tariff (NV*, β¬). The red dashed line denotes the lower bound of the water-CO2 performance, i.e., the existing electricity grids that have the lowest water consumption and CO2-eq emission simultaneously, and the green shaded area covers the power sources that can transcend the current limit of water-CO2 performance. |
651713f1ade1178b245e6e88 | 18 | Because electricity dominates the environmental impacts of LIB recycling processes, a comparison of electricity grid balancing areas that emit a range of CO2-eq emissions per MWh (averaged for 2019) are examined in Fig. (additional criteria air pollutants detailed in Table ). Substituting NEVP electricity with other balancing areas including Bonneville Power Administration Transmission (BPAT), California Independent System Operator (CISO), Western Area Power Administration of Colorado-Missouri (WACM), and a 100% renewable energy tariff in Nevada (NV*), yields a significant reduction in CO2-eq emissions of up to 93.3% (recycled scrap) and 87.4% (recycled battery) relative to conventional refinement (Fig. ). Conversely, employing low-carbon electricity grids can increase water consumption compared with NEVPbased operation, following the order of NV* > BPAT > WACM > CISO > NEVP (Fig. ). Note that NV*-and BPAT-based circular refinement processes exceed the water consumption level of conventional refinement due to significant contributions from hydro-and geothermal power. |
651713f1ade1178b245e6e88 | 19 | Further investigation into the grid electricity sources of balancing areas reveals a tradeoff between CO2-eq emissions and water consumption based on electricity generation type (Fig. ); most electricity sources with relatively low CO2-eq emissions (e.g., those based on bio-, hydro-, or geothermal energy) exhibit high water consumption, and vice versa. This tradeoff also explains the different influences of electricity source on environmental impacts of the Redwood Materials refinement step and other pathways (Fig. ). However, the electricity sources for each balancing area will affect both CO2-eq emissions and water consumption. For example, because NEVP-based electricity includes a relatively large proportion (70%) from CO2-eq emissions-intensive natural gas with low water consumption, a switch to hydro-intensive (73%) BPAT electricity decreases CO2-eq emissions while increasing water consumption. |
651713f1ade1178b245e6e88 | 20 | Environmental impacts of these upstream steps are analyzed for two representative LIB chemistries and battery use cases: NCA in EV battery packs, and lithium cobalt oxide (LiCoO2 or LCO) in smartphones. California is chosen to assess circular extraction because it has the largest population and EV market share of any state in the U.S. . Smartphones are considered extracted when collected, aggregated, and transported from all California residents (analyzed per census block) to the nearest existing collection facility (CF) . The analytical model for this circular extraction is depicted in Fig. , where a shortest-path route for collection from block group to CF is modeled . To quantify conventional material extraction environmental impacts from mining, global supply chain data are adapted from GREET (Table -S7) . Smartphone extraction in the circular supply chain emits only 0.0189 kg CO2-eq/kg LCO-eq, significantly lower than conventional mining (1.96 kg CO2-eq/kg LCO-eq) by 99.0%. Energy and water consumption are similarly lower in the circular supply chain (Table ). |
651713f1ade1178b245e6e88 | 21 | ). An algorithm is developed to quantify environmental impacts based on a weighted distribution of participating countries and the shortest distance along major transport routes (the case of cobalt is presented as an example in Fig. .) Conventional mine-to-refinery environmental impacts are calculated for one kg of embodied Li, Ni, Co, and Al metal (Table ). While transport emissions for Li, Ni, and Co range from 5.4-6.4 kg CO2-eq/kg embodied metal, Al is three times lower. For the circular case applied to California, smartphones and EV battery packs collected at CFs are transported to a hypothetical central LIB circular refinement facility at the population-weighted center (i.e., gravity point) of California (near Bakersfield) . In conventional supply chains, transporting mined material concentrates accounts for 3.68 kg CO2-eq/kg NCA-eq and 4.32 kg CO2-eq/kg LCO-eq. By comparison, emissions for the transport of aggregated end-of-life NCA EV battery packs (i.e., not disassembled) and LCO smartphone batteries (not separated from phones) to a circular refinement facility are 0.073 kg CO2-eq/kg NCA-eq and 0.47 kg CO2-eq/kg LCO-eq, 98.2% and 89.1% lower than transport of mined concentrate, respectively. The reduction in CO2-eq emissions is attributed to differences in elemental concentrations of transported materials and aggregate transport distance (e.g., a weighted average of 224 km for circular NCA-eq materials, and 57,600 km for conventional NCA-eq materials). |
651713f1ade1178b245e6e88 | 22 | The refinement step dominates environmental impacts of circular and conventional supply chains. Combining material extraction, transport, and refinement steps yields a cradle-to-gate comparison of the most differentiated steps of conventional and circular LIB supply chains for producing battery-grade cathode materials (Fig. ). Here the environmental impacts of the LIB refinement step in California are analyzed for a hypothetical scenario employing the same circular multi-step refinement technologies as Redwood Materials (i.e., RC+Me+Hy) in Nevada, but using California (CISO) electricity to produce battery-grade cathode materials. A circular supply chain in California for NCA EV and LCO smartphone batteries lowers energy and greenhouse gas emissions by at least 47.3% and water consumption by over 37.6%. In the case of recycling NCA EV batteries in California, the entire cradle-to-gate greenhouse gas emissions of the circular supply chain are lower than the transport emissions of mined concentrate in conventional supply chains (Fig. , Table ). Circular production of LCO-grade materials leads to higher environmental impacts than that of NCA-grade materials based on the mixed-stream feedstock composition processed by Redwood Materials. Overall, upstream steps (extraction and transport) contribute marginally to the total environmental impacts of both circular supply chains, accounting for β€4.9% CO2-eq emission, β€8.2% energy consumption, and β€0.24% water consumption. Accordingly, the refinement process dominates the environmental impacts of the circular supply chain. In contrast, upstream steps in the conventional supply chain play a larger role (still smaller than refinement) in cradle-to-gate environmental impacts, contributing between 7.8-31.0% to the environmental metrics considered (Table ). Environmental impacts of refinement are analyzed based on electricity generated from balancing grid authority CISO and upstream supply chain steps (extraction and transport) are based on data from GREET and transport models developed in the preceding section and depicted in Fig. . Specific environmental impacts of each step are detailed in Tables S5-S7. recycling facility. Practical LIB feedstock and refinement pathways are analyzed from recycling company (Redwood Materials) and modeling is employed to examine the environmental impacts of upstream material extraction and transport steps. The analysis reveals that refining end-of-life LIBs into battery-grade cathode materials exhibits lower environmental impacts than conventional refinement of mined materials, mixed salts products are more beneficial for circular refinement, and the source of input electricity is the principal factor governing circular refinement environmental impacts. Upstream circular supply chain steps contribute marginally to overall environmental impacts, and the refinement step comprises the largest source of cradle-to-gate environmental impacts. Disaggregated analysis of LIB refinement pathways at Redwood Materials provides important insights into the performance and potential of different refinement processes. While pyrometallurgical processing is widely considered as more environmentally intensive than hydrometallurgy, Redwood Materials' RC pathway exhibits much lower environmental impacts than current Hy-containing pathways reported in practice and in literature (Fig. ). The optimized conditions of RC processing minimizes the combustion of carbon-containing LIB materials, significantly reducing CO2-eq emissions while simultaneously generating products that are amenable for hydrometallurgical separation. Because chemical consumables such as H2O2 are important contributors to hydrometallurgy, environmental impacts of Hy processes could be reduced through more sustainable (e.g., electrochemical) production methods . Our findings also advocate the refinement products of mixed metal sulfates over the single salts, indicating that the further separations among Ni and Co salts can be avoided. An emerging alternative LIB recycling technology, "direct recycling", recovers functional battery materials without decomposition into substituent elements, and is reported to exhibit comparable environmental impacts to Redwood Materials methods . However, direct recycling is still under development and warrants further assessment after process optimization and industrial-scale implementation. |
651713f1ade1178b245e6e88 | 23 | Electricity greatly influences environmental impacts in LIB circular refinement, and the variability among grid electricity sources elucidates a tradeoff between CO2-eq emissions and water consumption (Fig. ). Therefore, considering water consumption and CO2-eq emissions is necessary for selecting recycling facility locations, particularly in water-sensitive or emissionssensitive scenarios. Further examination suggests that the tradeoff is primarily driven by waterintensive hydroelectric and geothermal electricity in certain locations versus CO2-intensive coal and natural gas in others, implying that increasing the proportion of electricity from nuclear, wind, and solar energy sources simultaneously reduces CO2-eq emissions and water consumption relative to existing balancing areas (Fig. ). |
651713f1ade1178b245e6e88 | 24 | Analyses of upstream environmental impacts inform better operations for future resourcesaving extraction and transport. Conventional mining and concentrating of ore or brine is resourceintensive due to the low natural concentrations of critical materials (0.01-1%), while critical material concentrations for transport rise to 3-15% after beneficiation. Further concentrating materials near mine sites or building reinterests closer to sources can efficiently reduce environmental impacts of the conventional mined materials. In contrast, smartphones contain 5% LCO material by mass, with the batteries themselves at 24% LCO . Circular material extraction via LIB collection decreases environmental impacts by 99% versus conventional. A "shortestroute" approach is used in this study to quantify the environmental impacts of battery extraction and transport supply chain steps. Practical battery collection operations will likely vary based on route selection and preprocessing strategy further influencing environmental impacts . For example, the disassembly of collected EV battery packs or removal of smartphone batteries from devices prior to transport to a recycling facility can increase energy usage through extraction but reduce environmental impacts by lowering transportation weight (Table ). Trucks are used as the primary vehicle for transport analysis given regulatory concerns that consider LIBs hazardous material in many transportation scenarios . However, alternative transport like railway can further lower environmental impacts by approximately four times versus trucking (Tables ). Upstream process optimization of environmental impacts warrants further investigation, such as the active area of high-throughput automation of LIB extraction from non-standardized devices and EV battery packs or rapid assessment of LIBS for second life uses. |
651713f1ade1178b245e6e88 | 25 | As the prevalence of LIBs grows in the mobility sector and beyond, strategic placement of domestic LIB collection, refinement and manufacturing facilities can further minimize future environmental impacts by considering heterogenous LIB growth by location, collection approach, transportation distance, and electricity source for refinement processes. As LIB production scales, policies informed by consumer surveys, focus groups, pilot testing, and diverse stakeholder engagement will be needed to research and scale battery collection . Business models for collection of all LIB types and sizes will likely vary from manufacturer-led to municipal or private collection programs. In addition to collection costs, the varied scale of collection requires further investigation, particularly for localized environmental impacts. Notably, analogous economic and environmental impacts to local ecosystems of conventional mining are not considered in this analysis, and warrant future studies . Additionally, designing and manufacturing LIBs for recycling in a circular economy can reduce resource usage identified in this study . Future efforts should also focus on optimizing refinement processes for subsequent steps of the circular supply chain in LIB manufacturing, product performance, and economic cost. |
651713f1ade1178b245e6e88 | 26 | Goal and scope. The goal of this study is to compare stepwise cradle-to-gate environmental impacts (energy consumption, CO2-eq emission, and water consumption) for two supply chains: a conventional, linear supply chain fed by natural mined material for refinement into battery materials, and a circular supply chain fed by LIBs. Both supply chains produce battery-grade cathode materials. A cradle-to-gate analysis of the whole supply chain considers steps of material extraction, transport, and refinement, and gate-to-gate analysis investigates the refinement step, which is focused on in this study. A gate-to-gate scope is broadly defined as the boundary surrounding processing facility operations. In this analysis, gate-to-gate refinement only considers direct processing (e.g., alteration, concentration, precipitation) of the feedstock material once it is extracted from its original state and transported to the refinement location (shown in Fig. ). For Redwood Materials, this scope includes mechanical processing, reductive calcination, and hydrometallurgy (Fig. ). The system boundary does not include other operations outside of the direct refinement processes as discussed in study limitations below. |
651713f1ade1178b245e6e88 | 27 | Two LIB feedstock streams are evaluated: (1) battery production scrap and (2) mixed, spent LIBs from consumers (Fig. ). The study scope upstream of the gate-to-gate supply chain step completes cradle-to-gate analysis, and includes both material extraction and transport steps. For conventional extraction, GREET is used for quantifying the environmental impacts of mining. |
651713f1ade1178b245e6e88 | 28 | Methodology. An attributional life cycle assessment is conducted to quantify and compare conventional and circular LIB supply chains for the production of battery cathode materials. This analysis complies with the International Organization for Standardization (ISO) 14040 standards but omits conversion to environmental impact indicators and external review ), normalizing by the mass of the individual element of interest within the output product (e.g., Li in Li2CO3) and then normalizing again by the mass of that element in the functional unit for this life cycle assessment (defined in the next section). For elements where more than one pathway of production exists in the GREET model (i.e., Ni and Li), the overall environmental impacts are calculated by averaging pathways weighted by their respective share of global production (45% Li production from brine and 55% from ore, and 60% Ni production from mixed hydroxide precipitate and 40% from Class Defining functional units. Functional units standardize comparisons of the resource consumption and emissions in life cycle assessments. In this study, two functional units are considered in this assessment to normalize environmental impacts between conventional and circular supply chains: the battery-grade material required to make one kg of stoichiometric lithium nickel cobalt aluminum oxide (LiNi0.80Co0.15Al0.05O2, NCA-eq) and lithium cobalt oxide (LiCoO2, LCO-eq) cathode material. Mass was selected as the primary normalizing factor because any energy-based functional unit (e.g., per kWh) could vary based on battery manufacturing and cycling characteristics. The NCA chemistry was selected because reports suggest future cathodes may utilize less Co compared to NMC batteries in EVs, and NCA comprised the second-largest category of EV battery chemistries in 2016, following NMC batteries . LCO is a representative chemistry used in handheld rechargeable devices (e.g., cellphones and laptops) which are currently available to recycle in larger quantities than EV LIBs. The environmental impacts of other LIBrelevant materials (Cu and Mn) in conventional supply chains can be found in Table . |
651713f1ade1178b245e6e88 | 29 | In both conventional and circular supply chains, the extraction, transport, and refinement steps are converted into environmental impacts metrics for the production of battery-grade materials and normalized by NCA and LCO functional units. A limiting reagent approach is used to quantify the environmental impacts of a functional unit in circular refinement pathways. |
651713f1ade1178b245e6e88 | 30 | Relatedly, Ni is the limiting output element from recycled batteries. For multi-step refinement processes, the recovery rate of Ni and Co is 95% and for Li is 92%. Additionally, a sensitivity analysis of environmental impacts from circular refinement is conducted based on facility location in different grid balancing areas and their associated electricity sources. |
651713f1ade1178b245e6e88 | 31 | (Table ). Three categories of environmental impacts are detailed in this study: energy consumption, air pollutant emissions, and water consumption. Energy consumption includes the input electricity for different applications and the energy required to produce required consumables. Criteria air pollutant emissions include the embodied emissions generated by the production of input electricity and the consumed reagents. CO2, CH4, CO, NOx, N2O, SOx, PM10, and PM2.5 are the air pollutants provided in the GREET model and considered here. The greenhouse gas emissions are reported as CO2 equivalents (CO2-eq) summing CO2, CH4, and N2O weighed by the corresponding 100-year global warming potential (GWP). Water consumption considers the withdrawn water that is not returned to the original source, and both the input city water usage and the embodied water consumption in electricity generation and the manufacturing of consumable materials are included. |
651713f1ade1178b245e6e88 | 32 | Estimating environmental impacts of material extraction. For conventionally mined ore and brine, energy consumption, CO2-eq emission, and water consumption values are separated for the material extraction processes found in the GREET model. For the circular extraction case, LCObased smartphones are assumed to be collected and transported to existing private and municipal collection facilities (CFs) from each census block group in CA, assuming every person owned a cell phone and purchased a new phone every three years. A shortest-route algorithm was used for refinement is excluded from the analysis due to the lack of information in GREET. In collection of end-of-life batteries in smartphones, inefficient transport to a CF (e.g., driving each smartphone individually or taking longer transport routes to a CF) is not considered. In addition, all end-of-life EV battery packs are assumed to be driven to each CF in their original vehicles, which is attributed to the "product use" stage instead of extraction in life cycle assessment; therefore, zero CO2-eq emissions are assumed for the extraction step of EV batteries. |
651713f1ade1178b245e6e88 | 33 | Transport between a domestic mine and refinery is not considered, resulting in net zero use of resources in such cases. The resources required to separate an embedded battery from its device prior to a refinement facility is not considered in a circular supply chains. Similarly, the effect of transporting only LIBs separated from the devices is not considered. Incorporating the domestic transport and separation operations can increase environmental impacts. |
651713f1ade1178b245e6e88 | 34 | Ancillary processes (e.g., transport between unit processes) beyond direct refinement unit processes and embodied resources of the capital equipment used for material refinement are not considered for the circular supply chain. The chemical formats of output products differ between the conventional and circular supply chains, but converting them to the same products will not substantially change the results due to the similarity between the cathode salts of the two supply chains (Note 3 in Supplementary Information). |
60c745ee842e653443db26f7 | 0 | This article assesses a key aspect of data sharing that has the potential to accelerate the progress and impact of medicinal chemistry. To achieve this the community needs to increase the outward flow of experimental results locked-up in millions of published PDFs into structured open databases that explicitly capture the connectivity between structures, documents and bioactivity results. But isn't there enough of this out there already? This can be answered in two parts. The first is that a conservative estimate of the capture backlog is that there is at least two-fold more data still entombed in PDFs not currently indexed in database records. The second part is the imperative to enable open science data mining at all scales. This applies not only to individual documents (small data) but scaling up to all papers and patents (big-data). The potential of the latter is huge, especially since Artificial Intelligence (AI) is being increasingly applied to knowledge distillation. This report will outline the principles of connectivity capture, sources, progress, impediments and prospects for their amelioration. |
60c745ee842e653443db26f7 | 1 | However, in the broader context of bioactive chemistry, it becomes indivisible from the related domains of chemical biology (directed towards mechanistic insight rather that direct drug discovery), enzymology, pharmacology, toxicology (as the wrong sort of bioactivity) in addition to the development of insecticides or herbicides ο· Connectivity: This term is used for an explicit link (e.g. a URL pointer) between a published document and the chemical structures specified in it. Implicit is not only manual navigation (e.g. link-clicking) but also that such connectivity can be made machine-readable and thus computationally interrogated at large scale via an Application Programming Interface (API) or a Resource Description Framework (RDF). |
60c745ee842e653443db26f7 | 2 | ο· Papers as documents: This typically refers to research papers from journals but increasingly needs to encompass their associated supplementary data . Note also that by far the majority of medicinal chemistry, biological chemistry and pharmacology papers are still behind subscription paywalls. However, the full-text for some of them is not only open but also available to be mined in both PubMed Central (PMC) and European PubMed Central (EPMC) . Connectivity can extend to other document types such as review articles and vendor catalogues. In this article the main document type referred to will be the PubMed Identifier . These have open abstracts and are also indexed in the Digital Object Identifier System (DOI). However, significant numbers of papers in the bioactive chemistry domain (including pre-prints) may be DOI-only. |
60c745ee842e653443db26f7 | 3 | ο· Patents as documents: Academics tend to overlook that patents a) include several fold more medicinal chemistry than papers b) appear years earlier c) most academic drug discovery operations apply for them d) they include a proportion of high quality data that never appears in journals e) they can be text-mined and d) consequently, over 23 million structures have entered PubChem via automated extraction. |
60c745ee842e653443db26f7 | 4 | ο· Non-document sources. While this article has to be restricted to documents it should also be noted that an increasing amount of drug discovery data is beginning to surface on the web that may never be instantiated in document form. Although this started with PubChem Bioassay as far back as 2004, the more recent proliferation is via Open-notebook science. More projects are using open Electronic Laboratory Notebooks (ELNs) that are not only accessible to anyone by web browsing, but also, crucially, crawled by Google and indexed for chemistry searching ο· Structures: A necessary focus of this article will be traditional small-molecule chemistry that is not too far outside the Rule-of-Five lead-like property space. In terms of connectivity value antibodies, other protein biotherapeutics, as well as large peptides or polynucleotides, are also important to encompass. However, capture into structured records is more challenging for these larger modalities than for smallmolecules that can be merged on the basis of chemistry rules. Notwithstanding, space limitations mean that non-small molecule connectivity is out of scope for this article. |
60c745ee842e653443db26f7 | 5 | ο· Open: As the theme of this special issue this term will doubtless be expanded on in other articles. However, brief qualification in the context of this work is necessary. Regardless of licensing complications, open is taken here to mean public data sources accessible via a web browser (signing in may be an impediment but not a stopper). These are thus distinct from commercial offerings where licenced access has to be purchased. |
60c745ee842e653443db26f7 | 6 | We can describe the connectivity between documents, structures and bioactivity as conceptual triage routinely performed by readers of medicinal chemistry or pharmacology papers . Typically, we peruse a document "D" that describes a bioactivity "A" with a quantitative result "R" (e.g. an IC50 assay) for chemical structure "C" that inhibits protein "P" in vitro (and by inference possibly in vivo). A useful shorthand thus becomes "D-A-R-C-P" (DARCP) as shown schematically in Fig. Figure . A schematic of the document > assay > result > compound > > protein target relationships, D-A-R-C-P |
60c745ee842e653443db26f7 | 7 | The order can be adapted to fit different use cases depending on the curation rules and data structures of the individual resources. For example, the substitution of "P" with target "T" can be used to indicate a cell or a microorganism, an SAR series can be represented as a multiplexed set of one-A to-many R-C. It can also be extended to "D-A-R-C-L-P" where L refers to the explicit location references for C in the document (e.g. "compound 10b" in a paper or "example 503" in a patent). However, it is limited as a formalism for bioactivity since there are many exceptions and mechanistic nuances that do not fit this simplification. A well-known example would be heparin (GtoPdb ligand 4214). In this case "C" can be a commercial partially purified extract of 12-15,00 Mw which consequently does not have a defined chemical structure. However, as a curatorial expedient, the defined 1040 Mw form (as PubChem CID 22833565) has been annotated, even though the sodium salt is the active form in vivo. Note also that while formally "P" is SERPINC1 (ATIII) the mechanism is an indirect one involving the activation of binding to F2 (activated thrombin) for inhibition. Another example of a shortcoming is mechanism-based covalent inhibition where the time dependence of IC50 is not captured. |
60c745ee842e653443db26f7 | 8 | ο· Documents: clustering by content relatedness, position within citation networks, connections via authors or institutional affiliations ο· Assays: classified by various assay ontologies ο· Results: log transformations (e.g. pIC50 or pKi) for potency sorting and implicit molecular mechanism of action (mmoa), (e.g. where A-R indicates C to be a potent inhibitor of P) ο· Compounds: a full range of cheminformatic analysis including 2D or 3D clustering, property prediction and chemical ontology assignments ο· Proteins: a full range of bioinformatic analysis including Gene Ontology (GO) assignments, pathway annotation, structural homology, disease associations and genetic variation (e.g. for target validation). |
60c745ee842e653443db26f7 | 9 | Those cases where the link is only compound-to-document can also be referred to as D-C-(or c2d). These have become available in a large excess over full DARCLP since they are technically easier to obtain and can be automated to a usable level of specificity. This needs the introduction of the intuitive concept of "aboutness" (ABNS). The title of a document from which D-C could be extracted usually includes an explicit ABNS statement. For example, the code name of a lead compound would be what a medicinal chemistry journal article would be "about". In the same way the ABNS of a clinical pharmacology journal article some years later could be describing the clinical trial results for the identical structure which, by then, could have an International Non-proprietary Name (INN). However, the extraction of multiple compounds (i.e. one-D-to-many-Cs) immediately becomes problematic in the absence of full relationship chains. For example, the med chem article may describe the testing of useful set of analogues for SAR but (as is usually the case) the A-R-P data was not extracted. |
60c745ee842e653443db26f7 | 10 | At this point we need to introduce the additional concept of name-to-structure (n2s). This is an important determinant of both ABNES and D-C utility. Using the example again from a paper this would mean that both the code name and the INN would be included in the D-C capture record (i.e. n2s) even if 50 analogues were also tested. Other examples that present particular ABNES problems are review articles, synthetic chemistry papers and patents. A review could exemplify 20 lead compounds all with different company code numbers and/or INNs, an extended synthesis report could give rise to 200 D-C records and a patent could have over 500. Discerning the ABNES for patents can be especially problematic since frankly obfuscatory titles and abstracts are common (e.g. "Novel Compounds") |
60c745ee842e653443db26f7 | 11 | This can be summarised by the following (unattributed) quote "We have spent millions putting data into the literature but now have to spend millions more getting it back out". This alludes to entombing the DARCP "meat" within a PDF "Hamburger". The paradox is that electronic text formats typically used for drafting papers are machine-readable (certainly with modern parsing techniques). However, this is systematically obviated by the PDF conversion. |
60c745ee842e653443db26f7 | 12 | For example, a chemist may have SMILEs and/or InChIs in their ELN and/or molfiles in an institutional data repository but have to convert this to a ChemDraw proprietary file format in order to render the structural image that eventually appears in the PDF. This means getting the structures "back out" for database capture needs either manual re-sketching or use of an image-to-structure (i2s) tool such as Optical Structure Recognition (OSRA) , both of which are error-prone processes. |
60c745ee842e653443db26f7 | 13 | The common practice of including tables of Markush representations (although they improve SAR readability), makes the extraction problem worse. While most medicinal chemistry journals will include IUPAC names in the synthesis descriptions, these also have to be pulled "back out" of the PDF. This can be done via PDF-to-text Optical Character Recognition (OCR) or curated by pasting across to the Open Parser for Systematic IUPAC Nomenclature (OPSIN) tool . Here again, both the automatic and manual procedures are error prone. |
60c745ee842e653443db26f7 | 14 | A specific example of the problem can be given for a 2017 Nature Disease Primer article on new antimalarial compounds entering development . Because the chemistry representations were restricted only to images in the PDF a blog post was necessary to manually map the structures to PubChem identifiers . The MyNCBI link to the 16 CID entries given at the top of the blog post are still live ( indicating reassuring persistence for this system after four years). While this initial connectivity was only D-C (and where D was a review article rather than a primary activity report) this example had an important sequel. |
60c745ee842e653443db26f7 | 15 | Since this report is about open connectivity it might not seem pertinent to review commercial resources. However, a brief review of these is relevant in several contexts. The first aspect to note is that, despite occasional use of the adjective "proprietary" in their descriptions, the primary content of commercial databases is almost entirely derived from open sources. Notwithstanding, they capture, curate, annotate, collate, integrate and index this in valueadded ways (including user-friendly, query front-ends and customer-specific APIs) to justify subscription costs. The second aspect is that by virtue of being able to apply more internal resources than open databases, their statistics indicate where the practical upper limits might lie. The third aspect is that they can give insights into the challenges of extraction, although technical details of how this is done are sparingly presented externally. |
60c745ee842e653443db26f7 | 16 | Consequently, this has to be classified as primarily D-C-only source. By November of 2019 SciFinder reached 157 million unique organic plus inorganic substances, having passed 100 million in June 2015. While some of these are virtual structures (i.e. never been synthesised) this large enterprise (with over 4,500 employees according to LinkedIN) has the de facto largest searchable collection of small-molecule structures extracted from papers, patents and other sources. A presentation from 2016 declared that in the first 7 months of that year ~10.5 million substances were extracted from ~0.5 million patents and ~1.0 million documents. In addition, ~ 75% of current novel structures are from patents . However, the 157 million is exceeded by the latest public UniChem release of just under 160 million . In addition, a 2019 scaffold diversity analysis stringently filtered the CAS collection down to only ~30 million compounds with direct links to literature and patents . Since its first release in 2009 Elsevier Reaxys has emerged as another large-scale D-C capture endeavour, the statistics and search characteristics of which have recently been compared with SciFinder . It has reached 31 million structures but also subsumes PubChem which brings it up to 105 million. |
60c745ee842e653443db26f7 | 17 | The two leading commercial sources that capture DARCLP at scale are Global Online Structure Activity Relationship Database (GOSTAR) from Excelra (formerly GVKBio) and Elsevier Reaxys Medicinal Chemistry . The current statistics for these are shown in Table . The GOSTAR numbers have a more detailed breakdown in a paper from 2013 (see Table in that reference) which includes the calculated curation averages of 12 compounds per-paper and 43 per-patent . Note that GOSTAR's compound total has doubled in the intervening six years but the extraction averages and ratio of compounds from papers : patents of ~ 1:2.7 recorded in 2013 are likely to be similar. Comparable detailed metrics for RMC curation have not been disclosed so it remains unclear what procedural differences that might explain their considerably larger activity, target and document counts compared to GOSTAR but connected to a million less compounds. Notwithstanding, using nominally the same medicinal chemistry corpus the extracted chemical structure ratios between SciFinder, Reaxys, GOSTAR and RMC are approximately 30:30:8:7. Several technical differences may explain these ratios but the most important is the primary focus of the latter two on full DARCLP capture rather than just D-C. This selectivity (for journals and patents) maintains the quality of target mapping and activity results. |
60c745ee842e653443db26f7 | 18 | The first web-instantiated curated resource, BindingDB, was published in 2001 . This was followed by the IUPHAR Ion Channels Compendium of papers in 2003. This had developed into the IUPHAR-DB website by 2009 and was updated to the current IUPHAR/BPS Guide to Pharmacology (GtoPdb) by 2012 . That same year also saw the first ChEMBL publication for which the website was live by 2010 . All three of these resources focus on expert-curated DARCP extractions from the literature. In addition, PubChem, first appearing in 2004 has now become the de facto global hub for DARCP because all the three databases above submit their structures that are integrated with ~700 other sources . |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.