id
stringlengths 24
24
| idx
int64 0
402
| paragraph
stringlengths 106
17.2k
|
|---|---|---|
60c73e16bdbb895770a37da2 | 12 | Due to its general nature, the SOAP kernel reports on intra-molecular correlations (essentially, the geometry of each isomer) as well as on the inter-molecular correlations (crystal packing and intermolecular contacts) between structures. The dual nature of the information encoded in SOAP-REMatch similarity leads to an increased intrinsic dimensionality of the landscape (as revealed by the decay of the eigenvalues of the kernel similarity matrix, shown in the supporting information) compared to that observed for single landscape analysis, where all crystal structures contain the same molecule. This higher effective dimensionality makes the choice of parameters of the kernel and of the sketch-map projection less straightforward. Previously, we found that using the hyperparameters which led to good lattice energy predictions from a Gaussian Process regression also led to insightful sketch-maps, although alternative choices of the environment cutoff distance emphasized different features in the crystal structures (close contacts vs long-range packing). The heterogeneous nature of the multi-landscape comparison exacerbates the dependence of the appearance of the map on the parameters of the kernel. This made it necessary to optimize the parameters in a trial-and-error fashion, which clearly makes the machine-learning analysis more biased than in the single landscape case. |
60c73e16bdbb895770a37da2 | 13 | The sketch-map that we obtain (Fig. ) includes all predicted structures of the 28 molecules. Isomer Group Index 0.12_4-6_4-6 With the parameters that we chose, the map captures the fact that different molecules can generate similar packing motifs; the 2D map features "islands" of structures that are populated by crystal structures from several different isomers (Fig. ). This finding is consistent with the concept of supramolecular synthons, according to which different molecules can adopt similar crystal packing motifs because they contain similar structure-directing functional groups: in this case, the same set of hydrogen bond donors and acceptors. |
60c73e16bdbb895770a37da2 | 14 | The clustering analysis confirms this finding; five high-dimensional HDBSCAN*-detected clusters are found across the landscapes (Fig. ), each comprising several different isomers. By comparing with Fig. , where the structures are coloured by the heuristic structure classification (Fig. ), it is clear that γ is the most common crystal packing adopted by the full set of molecules. This dominant cluster is also identified by the HDBSCAN* method, with good registry between the automatic clustering (cluster 3, dark blue in Fig. ) and heuristic classification (yellow cluster in Fig. ). The fact that automatically-detected structural motifs, as well as heuristic packing classes, comprise crystals formed by different molecules and also by molecules with different point-group symmetry suggest that both from the point of view of local environments (as detected by the SOAP-REMatch) and from the point of view of the stacking, there is no obvious relation between molecular symmetry and preferred packing. |
60c73e16bdbb895770a37da2 | 15 | Sheet-like crystal packing is less favored by most of the molecules. This packing type corresponds closely with cluster 1 from the clustering analysis (red cluster in Fig. ) and is mainly concentrated in the top/right-hand-side of the sketch-maps, which dominates the crystal pack-ing landscapes for molecules 5, 12 and 16. The crystal packing similarities among these three molecules could be largely attributed to the similarities in their molecular structures -within the molecule, hydrogen bond donors and acceptors are located close to each other on the convex side of the molecule. The herringbone packing motif is so infrequent across the crystal structure landscapes that it is not classed as a cluster by the HDBSCAN* approach. |
60c73e16bdbb895770a37da2 | 16 | While more work is needed to automate the definition of an effective representation of multilandscape maps, the possibility of representing simultaneously the crystal landscapes of multiple molecules in a way that highlights the presence of common structural features is already very useful as a guide to the interpretation of the interplay between molecular structure and supramolecular packing. |
60c73e16bdbb895770a37da2 | 17 | The crystal structures with the lowest (most negative) lattice energies mostly occupy the righthand-side of the sketch-map (Fig. ), a significant portion of which correlate to crystals formed by symmetric molecules. The trends in lattice energies between molecules show how the arrangement of functional groups can alter the overall strength of intermolecular interactions. However, in property-driven material design, it is the physical/chemical properties, in this case, charge mobilities, rather than the lattice energies, that are of most interest. Fig. reveals that, among all the low lattice energy crystal structures considered, a small number stand out with particularly high electron mobility; these crystal structures belong to molecule 5 and are located in the same region of the sketch-map as the lowest energy crystal structures. This coincidence of high mobility with energetic stability is promising for the experimental realization of these crystal structures. |
60c73e16bdbb895770a37da2 | 18 | Our primary motivation for this study is to assess whether any of the hypothetical molecular crystals of 2-28 offer improved properties in comparison to the reported molecule 1. For organic semiconductors, we judge the fitness of molecule by the calculated carrier mobility. CSP is key to this assessment by providing putative crystal structures of each isomer and the validation on 1 demonstrates that the structure prediction methods used here are reliable for this type of molecule. |
60c73e16bdbb895770a37da2 | 19 | An important aspect of the molecular assessment is that CSP provides a complete landscape of possible crystal structures for each molecule, which allows us to examine the charge carrier mobilities for all structures on these landscapes. We focus our analysis on the low energy region of the landscapes and have calculated mobilities for crystal structures within 7 kJ/mol of the global minimum crystal structure for each molecule; most known polymorph pairs are separated by 7 kJ/mol or less, so we take this as the experimentally accessible energy range on the CSP energy landscapes. |
60c73e16bdbb895770a37da2 | 20 | Having an ensemble of possible crystal structures for each molecule provides a wealth of information, but complicates the process of ranking the molecules according to the predicted properties because each molecule is associated with a range of possible properties, depending on which structure it adopts when crystallized. In our previous work, we considered five quantities for ranking a small set of polyaromatic molecules with potentially good electrical performance in their respective crystalline forms: |
60c73e16bdbb895770a37da2 | 21 | we treat this as a molecular property, which we calculate for the isolated molecule. A large λ -has a detrimental effect on the charge mobility due to the exponential decay of charge The electron mobilities for structures with low mobilities are underestimated due to numerical instabilities in the calculation of the transfer integrals, t. A possible correction is discussed in the supporting information. This does not affect the results for the best ranking molecules. transfer rate with respect to λ -. Thus, ignoring the effects of crystal packing, molecules with low λ -are favoured. |
60c73e16bdbb895770a37da2 | 22 | where µ i is the electron mobility in the i-th structure and ∆E i is its lattice energy difference with respect to the predicted global minimum for that molecule. β = 2.70 kJ/mol is a decay constant fitted to the probability of observing a pair of polymorphs with an energy difference of ∆E i . This quantity provides a measure of the collective likelihood that a given molecule will crystallize into a structure with a large carrier mobility. |
60c73e16bdbb895770a37da2 | 23 | The values of these five quantities for all 28 molecules investigated here are summarized in Table and a selection of the CSP landscapes and ESF maps are shown in Fig. (the rest are given in the supporting information). Since λ -is effectively a molecular property, its value can be used to rank molecules as potential organic semiconductors without the use of CSP to provide structure predictions or any other form of structural hypothesis. There is a wide range of values for the reorganization energy, from 0.200 (molecule 1) to 0.341 eV (molecule 14). Thus, this measure predicts molecule 1 to be the best molecule among the entire set, followed closely by 5 and 7, whose λ -are only 1 and 5 meV higher than that of 1, respectively. We observe no strong relationship between molecular symmetry and λ -. A low value for the latter often implies that the molecule possesses a stable anionic state, which can be achieved by maximally delocalizing an extra electron over the entire molecule. This in turn depends on the nodal structure of the molecular LUMO, which may not be simply determined from the molecular diagram. |
60c73e16bdbb895770a37da2 | 24 | The drawback of assessing molecules based on λ -, without considering crystal packing, is clear from the values of µ GM , the calculated electron mobility of the most likely crystal structure of each molecule. These values have a weak relationship with λ -; for example, molecule 7, which has one of the lowest values of λ -, has a very low predicted electron mobility in its global minimum energy predicted crystal structure. In contrast, some of the isomers with high λ -achieve good electron mobilities in their lowest energy crystal structures; for example, molecule 15 has µ GM 5.6 times greater than that of molecule 1, despite a 53 meV higher λ -. Thus, we find that, over this range of reorganization energies, strong electronic coupling due to favourable crystal packing can make up for a moderately high λ -while, conversely, a molecule with a promising, low λ -can lead to poor properties if it adopts an unfavourable crystal packing. As a result of the strong influence of crystal packing, molecule 1 is not in the top 10 molecules when ranked by µ GM , despite having the lowest λ -. Molecule 5 has the highest predicted electron mobility among the global lattice energy minimum crystal structures of all 28 isomers (µ GM = 2.706 cm 2 /Vs), eclipsing the calculated mobility of the observed (and global minimum predicted) crystal structure of 1 (µ e = 0.240 cm 2 /Vs). 5 is the only molecule for which the low energy part of the CSP landscape is dominated by sheet-like crystal packing (Fig. ). The extended hydrogenbonded flat ribbon structure in the predicted global minimum structure for 5 (Fig. ) promotes π-stacking along the out-of-plane direction. |
60c73e16bdbb895770a37da2 | 25 | Molecule 4 ranks second among the whole set in terms of µ GM , and the corresponding crystal structure is shown in Fig. . 4 adopts a similar γ-type packing as 1 (Fig. ) and a key structural feature that might contribute to such a high electron mobility is a significantly reduced π-stacking distance from 3.465 Å for 1 to 3.402 Å for 4. Molecule 4 is also exceptional because there are no other predicted crystal structures within the 7 kJ/mol lattice energy window of the global minimum. Such large energy gaps between structures are unusual on CSP landscapes and this gives a high confidence that the predicted, high µ e structure would be observed if this molecule was synthesized. Moving on to µ max and ∆E to assess the molecules, molecule 5 is again the lead candidate in the whole set. The crystal structure corresponding to µ max for 5 (Fig. ) possesses similar hydrogen-bonded ribbons as the predicted global minimum. µ max tells us the best property that can be achieved for each molecule if we were able to crystallize the correct structure from the low energy region of the crystal structure landscape. We find that structures with enhanced electron mobilities relative to the global energy minimum are predicted at elevated lattice energies for a significant portion of the screened molecules (Table , Fig. ); the global energy minimum crystal structure rarely gives the best electron mobility. This observation is consistent with our previous investigations on azapentacene molecules, where dimers within crystal structures with significant co-facial overlap (giving rise to largest t and µ) are often energetically penalized by relatively large exchange-repulsion interactions. |
60c73e16bdbb895770a37da2 | 26 | It is molecule 4 now that has the largest µ across the whole set. This originates from the high µ GM and the large energy gap between the global minimum and the second most stable structure on the predicted lattice energy landscape, leading to µ = µ max = µ GM . Molecule 5 is slightly less favourable according to µ because, despite very high µ GM and µ max , the landscape for 5 also contains many low energy crystal structures with low predicted electron mobilities (Fig. ). |
60c73e16bdbb895770a37da2 | 27 | The difference in ranking of the molecules between the single-crystal structure measures (µ max and µ GM ) and the landscape-averaged approach ( µ ) provides an interesting dilemma: structure prediction is more certain for 4 because there is only one low-energy predicted crystal structure, so we would be more confident in obtaining this structure with its high predicted µ e . Molecule 5, on the other hand, has a greater potential because of higher µ e in some of its predicted structures. However, a risk in targeting 5 for synthesis is the crowded energy landscape that also includes low-mobility structures and the possibility that one of these less promising crystal structures is adopted instead of the predicted global minimum or the structure with µ max . The choice of which to target synthetically, or whether to synthesize both candidates, depends on the associated synthetic difficulty, which we do not attempt to judge in the present work. |
60c73e16bdbb895770a37da2 | 28 | Within the broad context of computational screening of functional materials, the accuracies of the underlying computational methods are not the only scientific challenge but also the associated cost in terms of computational time required, which has a direct impact on the feasible size of chemical search space that can be comfortably explored. |
60c73e16bdbb895770a37da2 | 29 | In this work, the total cost of CSP for all 28 molecules was approximately 31.7 kilo CPU hours (corresponding to just under a week on 200 CPUs), whereas the cost of computing the charge transfer integrals for the 8786 molecular dimers needed for electron mobility calculations using DFT was approximately 43.9 kilo CPU hours. The computational cost associated with structure and property predictions is similar and the high cost in running DFT calculations for property prediction is immediately apparent. Ongoing efforts to accelerate these types of property calculations, using methods such as machine-learning of structure-property relationships, are required to further advance the field. |
60c73e16bdbb895770a37da2 | 30 | Crystal structure prediction has sometimes been seen as computationally demanding and slow, which would be unsuitable for screening purposes within a suitable time scale for material discovery. With advances in sampling methods, making use of efficient, accurate atom-atom force fields and utilizing parallel, high performance computing, this is no longer true for rigid molecules of the size studied here. In this work, we demonstrate the combination of CSP with charge mobility predictions to screen 28 isomeric molecules as potential organic semiconductors. These molecules were hypothesized as structural variants of the recently synthesized 1. We further combined a recently developed machine-learning method to construct, for the first time, a multi-landscape sketch-map to directly compare the crystal packing landscapes of a set of related molecules to reveal common structural classes among the landscapes and to correlate these classes with crystal properties. |
60c73e16bdbb895770a37da2 | 31 | Ongoing challenges still remain to make the most of the methods presented here in the process of the discovery of new materials. Some of the most pressing are (i) the selection of a diverse set of molecular candidates -in this work, the set of molecules has been constructed manually, based on chemical intuition -(ii) acceleration of the crystal structure prediction and property prediction calculations which, in this work, contribute almost equally to the overall computational cost, and (iii) the determination of a data-driven procedure to determine the best parameters for the sketch-map projection and the kernel construction to further automate the structural analysis across multiple crystal structure landscapes, revealing the relation between molecular structure, crystal packing and materials properties. |
60c73e16bdbb895770a37da2 | 32 | The variation of electron mobility within the crystal structure landscape of a given molecule, and between landscapes of related molecules, demonstrates the important influence of crystal packing on this property: improved intermolecular coupling can more than make up for a moderate increase in the molecular reorganization energy. Within the molecules included in our study, we find that the most commonly favored crystal packing across all molecules investigated here is the γ packing motif, which is predicted to be favored for molecules with different point-group symmetry and functional group substitution patterns. Furthermore, sketch-map and cluster analysis suggests that crystal structures from different molecules that favor the γ-type packing exhibit a high degree of structural similarity, which can be attributed to the abilities of these molecules to form similar short-ranged hydrogen-bond patterns. |
60c73e16bdbb895770a37da2 | 33 | With regards to the screening of electron mobilities, we find that a number of the hypothetical molecules are predicted to give higher electron mobilities than molecule 1. Molecules 4 and 5 are most promising among the 28 molecules that we considered. Molecule 5 has the second lowest reorganization energy and a crystal energy landscape that is dominated by sheet-like crystal packings, leading to very high electron mobility in some its predicted crystal structures, including its global lattice energy minimum and a higher energy crystal structure that gives the highest electron mobility across all 28 crystal structure landscapes. Molecule 4 is attractive because of high electron mobility in the global minimum energy crystal structure and a large energy gap to the next hypothetical crystal structure; the global energy minimum is highly likely to be observed in this case. |
60c73e16bdbb895770a37da2 | 34 | The study represents an important advance in the use of CSP for guiding materials discovery, demonstrating the ability to screen a moderately large set of molecules on a timescale that is shorter than what would be required for synthesis and testing of the set of candidates. We hope that this work stimulates more widespread use of these computational methods, as well as encouraging the synthesis and characterization of some of the promising hypothetical molecules that have been identified. |
67c858e5fa469535b90c4f9a | 0 | The Critical Assessment of Computational Hit Finding (CACHE) challenges are prospective benchmarking exercises modeled after CASP where computational chemists and data scientists use their methods to predict small-molecule ligands for a pre-defined protein target . But unlike CASP, CACHE challenges are prospective: predicted molecules are tested experimentally and all data shared publicly. The goal of CACHE is to delineate the state-of-the-art in computational hit discovery, an area poised for breakthroughs driven by advances in artificial intelligence (AI). The first CACHE challenge (CACHE #1), focused on the WDR domain of LRRK2, a Parkinson's disease target. An apo structure of the targeted domain was available in the protein data bank (PDB), but no ligand had been reported at the time. CACHE #1 reflected a highly dynamic and explorative field; a few weakly active molecules were discovered, indicating that significant progress remains to be seen . |
67c858e5fa469535b90c4f9a | 1 | In CACHE #2, computational teams were challenged to find drug-like ligands targeting the RNA-binding site of the SARS-CoV-2 helicase Nsp13, a site with bound fragments in the PDB (PDB codes 5RLH, 5RLZ, 5RML, and 5RMM) (Figure ). The reported fragments had no measurable binding affinity but highlighted putative interaction hotspots in the RNA binding site of Nsp13, which is one of the two most conserved sites in the coronavirus proteome and represents an attractive target for novel antivirals . Nucleic acid binding sites are typically highly polar and poorly druggable, but low micromolar ligands targeting the RNA sites of SNRNP200 and HCV NS3 (PDB 5URM and 4OKS, respectively) have been reported, supporting the idea that these sites can successfully be targeted by small molecules in some cases. Helicases are a clinically validated target class but are often recalcitrant to medicinal chemistry efforts due to the transient nature of their conformational states . As such, well-characterized small molecule ligands for Nsp13 would represent valuable chemical starting points for drug discovery. image formed by superimposing experimental structures of Nsp13 in complex with four fragments and in complex with RNA and ADP (blue and orange respectively; PDB code 7RDY ). CACHE #2 participants were asked to find ligands targeting the RNA-binding site occupied by fragments. Electrostatic potential coloring of the binding site, revealing the overall polar area, and bound fragments are depicted in the inset. |
67c858e5fa469535b90c4f9a | 2 | Here, we review the computational workflows and associated hit rates of the 23 teams who participated in CACHE #2. In an initial "hit identification" round (Round 1), each team selected up to 100 compounds from the Enamine catalog resulting in 1957 molecules that were procured and tested using Surface Plasmon Resonance (SPR), a direct biophysical binding assay. Each computational group was provided with experimental data on their respective compounds and asked to select up to 50 commercial analogs of their experimentally confirmed compounds of interest. The goal of this "hit expansion" round (Round 2) was to establish chemical series with multiple compounds experimentally confirmed to further build confidence in determining successful computational workflows. In parallel, all teams were asked to predict active molecules from the library composed of all Round 1 compounds collectively selected by all participants, a complementary evaluation mechanism where participants predict from the same library. |
67c858e5fa469535b90c4f9a | 3 | As in CACHE #1, the participating teams used a diverse array of workflows. Overall, hit rates were low compared with virtual screening results typically reported in the literature, with no clear benefit of using methods supplemented by machine-learning over purely physics-based methods. Nevertheless, 13 experimentally validated Nsp13-targeting chemical series (binding affinities ranging from 1 to 90 µM) were identified by 11 different teams, representing starting points for the development of chemical probes to explore the antiviral effect of Nsp13 inhibition. |
67c858e5fa469535b90c4f9a | 4 | The CACHE #2 competition targeting SARS-CoV-2 Nsp13 was initiated with applications due in September 2022. As specified in the CACHE roadmap 2 , an independent applications review committee (Table ) selected 25 participants for CACHE #2, based on the results of a double-blind peer review process where each applicant evaluates and rates five randomly selected applications. Twenty three out of the 25 selected teams submitted their computational predictions within the specified two-month timeframe. |
67c858e5fa469535b90c4f9a | 5 | The computational workflows represented diverse design strategies, techniques and tools (Figure ). Out of 23 teams, ten used neural networks to generate or evaluate compounds, eight used crystallized fragments in the PDB to guide their design, seven used molecular dynamics simulations to account for protein flexibility, four used free energy calculation and two quantum mechanics to refine their prediction. |
67c858e5fa469535b90c4f9a | 6 | For example, the Poda-Hoffer team (workflow 1448 -WF1448), adopted a conservative, purely physics-based but well-established screening pipeline where Glide (Schrodinger, New York, Inc.) was used to screen a large and diverse library, with pharmacophoric constraints, against a conformational ensemble extracted from fragment-bound Nsp13 crystal structures in the PDB, along with a few conformationally refined snapshots from quick molecular dynamics simulations. The output was refined with another scoring function (HYDE, BioSolveIT) after considering crystallographic water molecules from the system. Both computational and medicinal chemists visually inspected the top-scoring molecules to finalize the selection. |
67c858e5fa469535b90c4f9a | 7 | The Moretti-Meiler team (WF1414) implemented the challenge on Drug-it within the Fold-it platform , where citizen scientists use an online gaming interface to grow fragments bound to Nsp13 available in the PDB. After multiple rounds of chemical modification, the closest commercial analogs were re-docked with RosettaLigand and ranked based on neural network-predicted absolute binding free energies . Interestingly, these widely divergent workflows ended-up producing the two best Nsp13 binders. |
67c858e5fa469535b90c4f9a | 8 | In Round 1, each team was asked to select up to 100 in-stock or make-on-demand compounds from the Enamine catalog, leading to a collection of 1957 compounds quite evenly distributed between participants (61 to 97 compounds each, Figure ). Participants were also encouraged to use badapple () to filter out promiscuous compounds , though doing so was not mandatory. Overall, compounds displayed drug-like properties, as reflected by the distribution of their Lipinski descriptors (Figure ). While three of the four fragments crystallized in the RNA site of Nsp13 included a carboxylic acid attached to a ring, compounds were diverse, as illustrated by a pairwise distance matrix of Tanimoto distances based on ECFP4 Morgan fingerprints calculated with RDKit (Figure ). Chemical diversity was also observed within selections from each team, with rare exceptions, outlined by darker squares along the diagonal of the distance matrix. |
67c858e5fa469535b90c4f9a | 9 | Only 20 compound pairs selected from different participants had a Tanimoto distance of 0.3 or lower, based on ECFP4 fingerprints. Not surprisingly, all closest analogs selected by different participants (Figure ) were also close analogs of the crystallized fragments found in the PDB (Figure ), however none of these were ultimately confirmed experimentally. Indeed, in the previously reported fragment screen by crystallography, Nsp13 crystals were soaked in 50 mM fragments solutions , which can lead to the capture of fragments that are too weak to be detected by SPR (maximum concentration of 200 µM). Yet, crystallographically captured fragments were successfully grown into 20-40 µM hits, as detailed below. |
67c858e5fa469535b90c4f9a | 10 | Helicases are complex and structurally dynamic enzymes that couple ATP (or other nucleotides) hydrolysis at one site with RNA or DNA duplex unwinding at another. Given that the fragments in the targeted Nsp13 structure (PDB codes 5RLH, 5RLZ, 5RML, and 5RMM) bound to full-length Nsp13 in the absence of ATP or RNA , a similar form of the protein was used in a surface plasmon resonance (SPR) assay to measure the direct binding of the 1957 Round 1 compounds to the full-length protein (Table ). Nsp13 is a core component of the replication-transcription complex that also includes the viral RNA-dependent RNA polymerase (RdRp) , but the isolated protein was used in the assay for two reasons: first, fragments in the PDB were bound to the isolated monomer, and second, binding to RdRp would have obscured the results. All compounds were also tested in an ATPase assay (Table ), but we saw no correlation between SPR and ATPase assays and decided to rely on direct binding (SPR) to advance compounds to Round 2. Indeed, false positives in the ATPase assay that may bind to other assay-specific molecular components should be true negative in SPR, while true positives binding the RNA site in the SPR assay may not inhibit the ATPase activity. We also cannot discount the possibility that SPR hits may bind at unexpected and functionally neutral sites. |
67c858e5fa469535b90c4f9a | 11 | All compounds were tested at 50 µM in both assays. 300 compounds had acceptable SPR sensorgram profiles with a binding signal above 50% of the expected signal (based on the amount of protein captured on the SPR streptavidin chip), and were advanced to dose-response by SPR. Another 54 compounds that inhibited the ATPase activity by 40% or more at 50 µM were selected for SPR dose-response. Dose response measurements were conducted on the resulting 354 compounds by SPR, as well as on 96 compounds in the ATPase assay. Binding to WDR5, an unrelated protein, was also measured by SPR for selected hits to flag non-specific binders. Compounds of interest with signs of poor solubility or aggregation (<80% detected laser power at 100 µM) as measured by dynamic light scattering (DLS) were also flagged but were not dismissed to avoid false negatives (Table ). Indeed, unlike a typical drug discovery program, no active compound should be left behind in CACHE, as this would defeat the purpose of evaluating the efficiency of computational predictions. In the end, 46 compounds selected by 18 teams had a K D < 150 µM, a binding signal between 30% and 150% of the expected signal, and were advanced to Round 2 (Figure , Table ). While most hit rates were between 0 and 3%, workflows WF1454, WF1418 and WF1456 had significantly higher hit rates (9%, 8% and 7% respectively). The overall Round 1 hit rate was 2.3%. |
67c858e5fa469535b90c4f9a | 12 | The goal of the second round was to build confidence in advanced hits by experimentally verifying that their chemical analogs were also binding to the target. Compounds associated with experimental orange flags, such as signs of aggregation or poor solubility, were advanced to Round 2 to avoid false negatives and unfairly discounting computational methods. Seventeen teams selected up to 50 analogs of their Round 1 compounds of interest (compounds showing a binding signal by SPR), leading to 618 Round 2 molecules that were screened at 50 µM in an SPR binding assay, followed by dose-response and measurements of aggregation and solubility, as in Round 1 (Tables ). Compounds were also tested in an ATPase assay (Tables and respectively), and no correlation was observed with SPR data, as in Round 1. F-NMR was used as an orthogonal binding assay for fluorinated molecules. |
67c858e5fa469535b90c4f9a | 13 | Multiple chemical series emerged from this exercise (Table , Figure , and Supplementary material). Thirteen of the high ranked compounds as well as compound derivatives that scored lower but exhibited high binding affinity by SPR were further assayed for inhibition of double-stranded RNA unwinding activity in a FRET-based assay. Two compounds were found to potently inhibit Nsp13 helicase activity (CACHE2-HO_1431_6: K D 770 nM ± 180, unwinding IC 50 8.6 µM ± 1.7; CACHE2-HO_1454_15: K D 31 µM ± 0.7, unwinding IC 50 57 µM ± 2). Inhibition of dsRNA unwinding by CACHE2-HO_1431_6 was also confirmed in a gel-based unwinding assay with CACHE2-HO_1454_15 partially inhibitory, consistent with CACHE2-HO_1431_6 having a more potent unwinding activity and stronger binding affinity and lower IC 50 value. (Figure ). Note that many compounds in this series have an ester group linker that is likely to be hydrolyzed in cells and represents a serious medicinal chemistry liability, which penalized the final score of this chemical series. However, modifying the ester linker to a more stable group could easily address this liability while conserving potency. But this medicinal chemistry work is beyond the scope of the CACHE study. |
67c858e5fa469535b90c4f9a | 14 | The biophysical data and structure-activity relationship (SAR) of Round 1 hits and their Round 2 analogs were evaluated by an independent Hit Evaluation Committee composed of industry experts in biophysics, medicinal chemistry and computational chemistry (Table ), leading to a final score assigned to each Round 1 hit (Table ). Overall, 13 compounds had a score greater than 10 (Table ), reflecting robust experimental confirmation, which corresponds to a hit rate of 0.7%. The computational workflows of CACHE #2 participants were then evaluated based on the aggregated score of Round 1 compounds, and based on the best scoring Round 1 molecule (Figure ,b, Table ). In a separate evaluation scheme, all participants were asked to predict Nsp13 ligands from the merged collection of 1957 Round 1 compounds before the experimental data were generated. The aggregated score of predicted hits, normalized based on the number of hits predicted, was used to rate the computational workflows (Figure ). This scheme is complementary as here, all teams predicted hits from the same library, while in Rounds 1 and 2, participants screened compound collections from the Enamine catalog that may vary widely in size to best align with their computational methods and resources. ). |
67c858e5fa469535b90c4f9a | 15 | While these combined metrics provide a complete evaluation of computational workflows used in CACHE #2, a list of six well-performing workflows was compiled for further analysis, including WF1454 and WF1456, which had the best two aggregated scores, WF1414, WF1448, and WF1419 which predicted the three best scoring chemical series, and WF1438 that did best in predicting hits out of the 1957 Round 1 compounds (Figure ). Importantly, absence from this selection focused on top-performing computational pipelines does not imply that a workflow failed. |
67c858e5fa469535b90c4f9a | 16 | Most of the best scoring compounds were docked to the RNA binding groove, at the site occupied by fragments found in PDB structures 5RMM, 5RLZ and 5RLH, the target site defined for this CACHE challenge (Figure ). An exception is CACHE _1454_98, which is predicted to occupy an unrelated binding pocket. In this workflow (WF1454), compounds were docked onto a receptor grid spanning most of the target protein. Six of the eight Round 1 compounds from WF1454 that advanced to Round 2 occupied the RNA-binding groove, one the ATP site and one (CACHE _1454_98) an unrelated site. While top hits from other workflows occupy the expected site, they do not share pharmacophoric features or conserved interactions. CACHE_1414_40 was obtained from growing the crystallized fragment found in the PDB structure 5RMM (Figure ) and is predicted to loosely overlap with the bound fragment. The six best-performing workflows (Figure ) can be divided into three groups (Figure ). WF1414 and WF1438 both adopted strategies where fragments from the PDB were gradually grown and commercial analogs identified along multiple iterative cycles but their implementations were drastically divergent: WF1414 relied on citizen scientists and the gaming interface provided by Foldit to grow fragments, followed by RosettaLigand , a physics-based docking tool, and BCL-AffinityNet, a feed-forward deep neural network, for final scoring ; WF1438 used FEgrow to enumerate fragments in the binding pocket based on a hybrid machine learning (ML) / molecular mechanics energy function leveraging the ANI neural network potential for ligand energetics, and final evaluation with the convolutional neural network scoring function GNINA . |
67c858e5fa469535b90c4f9a | 17 | Another selection strategy adopted in workflows WF1456 and WF1454 was to dock a small and diverse library with GNINA or Vina 20, respectively, to initiate iterative active learning cycles where a ML model is trained on a small set of docking scores to predict ML-scores for billions of commercial compounds, and where ML-scores are used to select the next small subset for docking and refinement of the ML model. In WF1454, the selection was further refined with a round of consensus scoring. |
67c858e5fa469535b90c4f9a | 18 | Finally, WF1448 and WF1419 implemented a more direct approach where a large and diverse library was docked with pharmacophore constraints followed by orthogonal re-scoring. WF1448 used purely physics-based approaches for docking (Glide) and Scoring (HYDE), followed by visual inspection and selection of top compounds by both computational and medicinal chemists. WF1419 used the popular open-source software Vina for docking combined with ML/deep learning scoring functions RF-Score-VS and SCORCHs . |
67c858e5fa469535b90c4f9a | 19 | Overall, five of the top six performing workflows combined physics-based and ML techniques. All five workflows used ML to score docked poses, and two (WF1454 and WF1456) used ML to accelerate screening within active learning cycles. A more conventional, purely physics-based approach (WF1448) also proved successful, demonstrating that well-established physics-based virtual screening techniques remain competitive when deployed by experienced computational chemists. While only 22% of the workflows (five out of 23) used in Round 1 explicitly accounted for protein flexibility using conformational ensembles (WF1419, WF1422, WF1447, WF1448) or flexible docking (WF1414) (Figure ), they represented 50% of the most successful workflows (three out of six) (Figure ). Considering the well-known conformational dynamics of helicases , including Nsp13 5 , accounting for receptor flexibility may indeed have increased chances of success. |
67c858e5fa469535b90c4f9a | 20 | In CACHE #2, computational teams were asked to find molecules that target a pocket occupied by fragments in the PDB, a common challenge successfully met by the COVID moonshot initiative that targeted the SARS-CoV-2 protease , which could also be undertaken for other targets. In our challenge, the crystallized fragments were weak and had no measurable binding affinity by SPR (data not shown). Only eight out of 23 computational teams explicitly used the bound fragments in their selection strategy, and two of these were among the most successful workflows (Figures and). This shows that rationally optimizing crystallized fragments remains a challenging exercise that requires further developments before it can be reliably applied. Considering the multitude of targets with bound fragments in the PDB, including those taken to fragment screening by crystallography , technological development in this area of computational design could be impactful. |
67c858e5fa469535b90c4f9a | 21 | A main goal of CACHE is to highlight computational strategies that repeatedly perform well within a challenge or across multiple targets. Interestingly, using physics-based docking data on a relatively small library to train a ML model that can then be used to efficiently navigate a much larger chemistry space was a winning strategy both in CACHE #1 (WF1193 and WF1209) and CACHE #2 (WF1454 andWF1456) (Figure ). Among the dozens of commercial and open-source computational tools used by CACHE participants, the convolutional network scoring function implemented in GNINA was found in one winning workflow in CACHE #1 (WF1181) and in two in CACHE #2 (WF1438 and WF1456), strongly suggesting that this software is robust across two targets absent from training sets (no ligand with measurable binding affinity was previously known for either CACHE target). Fragment-based techniques linking docked fragments in CACHE #1 (WF1183 and WF1202) or growing crystallized fragments in CACHE #2 (WF1414, WF1421, and WF1438) also define a recurrently successful approach to computational ligand design. Workflow WF1414 is a distinct variation on this theme in that it relies on the design of citizen scientists who use a gaming interface to grow fragments in a binding pocket after which designs are evaluated with RosettaLigand. Combining human creativity with tools such as RosettaLigand may indeed be a recipe for success. |
67c858e5fa469535b90c4f9a | 22 | Only one of the CACHE participants explicitly included the visual inspection and subjective judgment of medicinal chemists as a final step in their selection strategy (WF1448). This step is common practice in virtual screening and should be better tracked in the future. Indeed, in its current set-up, CACHE evaluates not only computational methods but also the intuition and expertise of humans running these tools. The most seasoned computational chemist will be hard pressed to subjectively select hits out of a failed computational workflow. We would therefore argue that experimentally confirmed hits can only reflect successful computational workflows. Nevertheless, there would be some merit in requesting a more detailed description of human intervention from CACHE participants, including asking them to provide "computer-only" selections in addition to their final, human-selected sets (if any), at the risk of spending resources on testing compounds that do not pass the subjective evaluation of experts. |
67c858e5fa469535b90c4f9a | 23 | We observed a significant improvement in binding affinity for only one chemical series where two analogs (CACHE2-HO_1421_29 and CACHE2-HO_1421_27) showed a 3-fold increase in binding affinity compared with the Round 1 hit (CACHE_1421_62). The other exception is CACHE2-HO_1431_6 but it is a very distant analog of the parent molecule. The limited improvement seen in Round 2 may reflect a limitation in the commercial availability of analogs. Indeed, dedicated chemistry is typically preferred for the design of highly customized molecules. A mechanism to mitigate this effect could be to focus Round 1 screening on compounds richly derivatized in commercial catalogs. We expect that such an approach will become more attractive in the future, as commercial libraries keep growing. |
67c858e5fa469535b90c4f9a | 24 | Retrospective benchmarking exercises are critical to compare predictive computational methods and carefully assembled datasets play a central role for example to evaluate docking, virtual screening or free energy perturbation methodologies . While the value of these resources is generally well appreciated among computational chemists and data scientists, one may be surprised to see new ML-driven virtual screening tools being published every month that perform better than "all others" when tested for example on the PoseBusters dataset . Skeptical data scientists may wonder whether data leaked between training, test and validation sets while seasoned drug-hunters and experimentalists may refer to the old Danish proverb saying that "It is difficult to make predictions, especially about the future". |
67c858e5fa469535b90c4f9a | 25 | In CACHE #2, 23 computational teams were challenged to prospectively predict ligands for the RNA binding site of SARS-CoV-2 Nsp13, a binding pocket with no known drug-like ligand. Testing the predicted compounds experimentally yielded a low hit rate of 0.7 %, indicating that a breakthrough in computational hit finding where bioactive molecules are reliably designed in silico remains to be seen. Strikingly, the highest scoring prediction in CACHE #2 was a compound manually designed by citizen scientists using the Fold-it online interface and further prioritized by physics and ML-based computational tools (WF1414), emphasizing the value of human intervention in the design process. Computational hit finding strategies and tools recurrently successful across the first two CACHE challenges define emerging trends that may inform the community when constructing hit-finding computational pipelines. To the best of our knowledge, the thirteen compounds confirmed experimentally are the first with a measurable binding affinity expected to engage the RNA binding site of Nsp13. Considering the exceptionally high conservation of this site and its central role in the essential replication-transcription complex , molecules discovered in CACHE #2 provide valuable chemical starting points for future medicinal chemistry exploration. |
67c858e5fa469535b90c4f9a | 26 | DNA fragments encoding SARS-CoV-2 Nsp13 residues A5325-G5925 were amplified via PCR and sub-cloned into the pFBD-BirA expression vector. The insert was positioned downstream of the AviTag for in vivo biotinylation and upstream of a HisTag. The resulting plasmid was transformed into DH10Bac™ competent E. coli (Invitrogen) and a recombinant viral bacmid DNA was purified and followed by a recombinant baculovirus generation for baculovirus mediated protein production in Sf9 insect cells. Biotin was added to the medium at a final concentration of 10 μg/mL. Cells were harvested by centrifugation at low speed (2500 rpm for 10 minutes at 4 °C in a Beckman Coulter centrifuge) when cell viability dropped to 70-80%. The cells were resuspended in extraction buffer (20 mM Tris-HCI, pH 7.2, 500 mM NaCl, 5% glycerol, 5 mM Imidazole + 1 ml PI cocktail (Aprotinin, Leupeptin, Pepstatin A, and E-64) and lysed chemically by adding NP40 (final concentration of 0.5%) and 5 µl/L Benzonase Nuclease (in-house) followed by sonication at the frequency of 7.0 kHz (5" on/17" off) for 3 min (Sonicator 3000, Misoni). The crude extract was then clarified by high-speed centrifugation (60 min at 36,000 ×g at 4 ˚C) in a Beckman Coulter centrifuge to remove the cellular debris. The clarified lysate was first sent through a Ni-NTA resin column followed by passage through Gel filtration HiLoadTM 26/600 Superdex (Cytiva) with 50 mM Tris, pH 7.2, 200 mM NaCl, 5% glycerol, 0.5 mM TCEP to enrich nsp13_SARS2 to 95% purity. Following the identification of the protein eluting fraction and purity using SDS-PAGE gels, and mass confirmation, the fractions were pooled, concentrated, snap-frozen, and stored at -80 0 C until use. Protein mass was confirmed by LC-MS. |
67c858e5fa469535b90c4f9a | 27 | The assay was conducted using a Biacore™ 8K (Cytiva) at 20 °C. Biotinylated Nsp13_SARS2, with approximately 4900-5100 response units (RU), was immobilized onto the flow cell two of a streptavidin-conjugated streptavidin chip following the manufacturer's protocol. The flow cell one served as a reference for subtraction for each channel. Compounds were initially dissolved in 100% DMSO to create 10 mM stock solutions, which were subsequently serially diluted (factor: 0.5) to obtain six concentration points in 100% DMSO. For the SPR run, these serially titrated compound stocks were diluted at the ratio 1:50 in HBS buffer, containing Mg2+ (10 mM HEPES pH 7.4, 150 mM NaCl, 5 mM MgCl2, 0.03% (v/v) Tween 20) to achieve a final DMSO concentration 2%. Binding experiments used multi-cycle kinetics with a contact time of 60 seconds and a dissociation time of 180 seconds at a flow rate of 40 µL/min at 20 °C. The dissociation constant (KD) values were determined using steady-state affinity 1:1 binding with the Biacore™ Insight Evaluation software (Cytiva). |
67c858e5fa469535b90c4f9a | 28 | The solubility of compounds was estimated by DLS that directly measures compound aggregates and laser power in solution. Compounds were serially diluted directly from DMSO stocks, then diluted 50x into filtered 10 mM Hepes pH7.4, 150 mM NaCl, 5 mM MgCl2, 0.03% Tween20 (2% DMSO final). The resulting samples were then distributed into 384-well plates (black with a clear bottom, Corning 3540), with 20 μL in each well. The sample plate was centrifuged at 3500 rpm for 5 min before loading into DynaPro DLS Plate Reader III (Wyatt Technology). |
67c858e5fa469535b90c4f9a | 29 | The level of ATP consumed by Nsp13 was quantified by measuring the amount of remaining ATP using a luciferase-based assay as previously described . The inhibitory effects of compounds were assessed in 384-well format (14 μL final volume) using reactions composed of 50 mM HEPES, pH 7.5, 5% Glycerol, 5 mM magnesium acetate, 5 mM DTT, 0.01% Triton X-100, 0.01% BSA, 0.1 nM Nsp13, 3.5 nM 30b PolyT ssDNA, 2.5 µM ATP, and 2% DMSO. Samples containing DMSO only (no compounds) were used as a control. Reactions were started by the addition of substrate and incubated for 60 min at room temperature. Then, 10 μL of the reactions were transferred into 384-well white plates containing 10 μL luciferase reagent (Cat# V6712; Promega, Madison, WI, USA) and incubated for another 20 min at room temperature. Compounds that were followed up for dose-response experiments were tested using the same luciferase reagent, and the data were analyzed using GraphPad Prism 9. |
67c858e5fa469535b90c4f9a | 30 | The binding of fluorinated compounds was assayed by assessing the broadening and/or perturbation of F resonances upon addition of Nsp13 (at protein to compound ratios of 2:1 to 3:1) in PBS buffer (pH 7.4, 137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4, and with 5% D2O). 1D- F spectra were collected at 298 K on a Bruker AvanceIII spectrometer, operating at 600 MHz, and equipped with a QCI probe. Two to four thousand transients were collected with an acquisition period of 0.2 s, over a sweep width of 150 ppm, a relaxation delay of 1.5 s, and using 90° pulses centered at -120 ppm. The concentration of the compounds in both reference and protein-compound mixtures was10 μM. TFA (20 μM) was added as an internal standard for referencing. Prior to Fourier transformation, an exponential window function was applied (lb = 1 to 3) to the FID. All processing was performed at the workstation using the software Topspin 3.5. |
63c835ea5ab3135f8eac3f4a | 0 | For decades, cyclic imine (CI) toxins have stimulated extensive interest from the broad scientific community based on their unique and potent bioactivity coupled with their captivating chemical structures . The therapeutic potentials of larger members in this family, such as pinnatoxins , spirolides (4), and gymnodimines , have been thwarted by their high neurotoxicity in vivo. More compact members of this family were isolated from benthic dinoflagellate Vulcanodinium rugosum in 2013 and 2018, portimine A (PA, 1) and B (PB, 2) respectively (Fig. , absolute configuration confirmed in 2019) . In sharp contrast to classic CI toxins, preliminary reports indicate that 1 is highly cytotoxic (~3 nM) and induces apoptosis in several cancer cell lines, while also displaying lower acute toxicity in mice relative to other shellfish toxins, highlighting its potential as a therapeutic agent . However, the mechanism of action (MoA) behind its potent activity is unknown. The Achilles heel of such a compound is of course scalable access to its complex architecture. As it is derived from a dinoflagellate in low yield , chemical synthesis appears to be the only means of realistically procuring such molecules. Even if a bioengineered synthesis could be achieved, semi-synthetic analogs with deep-seated modifications would be unworkable. Featuring a spiro-fused five-membered cyclic imine embedded in a highly oxidized all-carbon tricyclic macrocyclic core, 1 and 2 are formidable targets for synthesis. The unusual peripheral oxidations such as that adjacent to imine carbon (C-5) and neighboring labile medium-sized cyclic ketal add to this challenge. In this work, the first total synthesis of 1 and 2 is presented, which features a carefully choreographed sequence to rapidly build up a minimally oxidized carbon framework followed by strategic oxidations and ring-chain isomerizations that minimize concession steps. In addition, this practical synthesis enabled elaborated in situ investigations as well as illumination of the cellular targets of 1 via chemical proteomics, revealing that 1 targets the 60S ribosomal export protein NMD3 and blocks polysome formation. |
63c835ea5ab3135f8eac3f4a | 1 | Historically, CI toxins have been constructed by patterning the retrosynthetic analysis on the presumed biogenesis , wherein an acyclic structure with maximum functionality is subjected to macrocyclization . This pioneering approach was first accomplished by Kishi et al. in the 1998 total synthesis of pinnatoxin A . In the case of the portimines, approaches thus far have followed this dogma (21-24). Thus, Brimble et al. and Harran et al. (24) aimed for a bio-inspired synthesis that mimics the polyketide synthases (PKSs), featuring bold intramolecular cyclizations of densely functionalized polyketides 3 and 4 respectively (Fig ). The former approach demonstrated that the ketalization of linear 3 was not facile, even with the well-functionalized skeleton. The latter cycloaddition-based approach resulted in undesired regioselectivity in the pivotal cyclization of 4 to 6. The difficulty encountered in these routes points to the challenge of forging key bonds in such a densely functionalized polycyclic alkaloid from an acyclic precursor. From a high level, this scenario is not unlike that encountered in the synthesis of densely functionalized, highly oxidized terpene natural products. In those cases, it has been shown that a two-phase approach to synthesis can be beneficial by building up a minimally oxidized carbon framework followed by strategic late-stage oxidations . |
63c835ea5ab3135f8eac3f4a | 2 | By analogy to the logic of two-phase synthesis, a minimally oxidized macrocyclic intermediate to the portimines was targeted with the assumption that a proper choreography of oxidation events would solve both connectivity and stereochemical issues. The only C-O bonds to be installed at the outset were those residing at C-4 (imine carbon) and C-10 (secondary alcohol). A triple bond on C-7/8 would be a surrogate for the eventual C-7 oxidation and, critically, offer a strategic disconnection to the macrocycle using ringclosing alkyne metathesis (RCAM) . Such a tactic would thereby minimize unstable functional groups and redundant redox manipulations since four key oxidations (C-5, C-13, C-14, C-15) would occur postmacrocyclization. To minimize protecting groups (PGs), the innate reactivity and conformational preferences are utilized via ring-chain reorganization/reconstitution steps. Upon unraveling the macrocycle, the dialkynated precursor traces back to accessible building blocks (Fig. ). Finally, in order to maximize access to useful analogs, a vinyl triflate was selected as a key functionality to be carried through the entire synthesis. |
63c835ea5ab3135f8eac3f4a | 3 | The synthesis is outlined in Fig. and commences with a scalable, asymmetric Diels-Alder cycloaddition , which established the C-3 chirality. Reduction of C-4, followed by removing the carbamate auxiliary with TBAF, afforded 10 in 88% yield, and 94% ee on 50-gram scale. The requisite methyl-capped alkyne side chain can be installed via a sequence including Grignard addition and two oxidations, delivering 11 as the final product in 60% yield after a single purification. Treatment of 11 with TFA in CH2Cl2 afforded spirocyclic imine 12 in 72% yield on gram scale. |
63c835ea5ab3135f8eac3f4a | 4 | The synthesis of fragment 13 was carried out from inexpensive (S)-solketal (ca. $1.1/g, see the SI for details). To affix this subunit onto the established chiral spirocyclic core 12, a stereoselective Cu-mediated conjugate addition was applied. The choice of copper (I) reagent was crucial since switching to other common copper (I) salts, such as CuI, CuCN, and CuBr, showed little to no observable conversion (Table ). The stereochemical outcome in this step is controlled by the intrinsic configuration of the spiro-cycle, wherein the side chain blocked the top face. Direct treatment of the in situ generated enolate with Comins' reagent ensured the correct regiochemical olefin outcome and smoothly delivered vinyl triflate 7 as a single product (6.5-g scale), which possessed all skeletal carbon atoms required for 1 and 2. |
63c835ea5ab3135f8eac3f4a | 5 | Attention was then turned to constructing the 14-membered macrocycle in portimines' skeleton through RCAM, a maneuver that might be derailed by the imine, olefins, or vinyl triflate. Fürstner's extremely efficient, canopy-shaped catalyst, [Mo] was chosen at this point due to its outstanding functional compatibility and demonstrated robustness . The pivotal RCAM step could indeed be achieved in 53-65% yield when heated 7 with [Mo] in toluene. However, a relatively high catalyst loading (12.5 mol %) was required, presumably due to the basic imine nitrogen. To lower the catalyst loading, the imine was masked with a Troc group, followed by exposing the formed enamide to 2.0 mol % . In this case, formation of macrocycle 14 was completed in one hour. Treating crude 14 with acidic wet methanol liberated the C-4 ketone and deprotected the TBS ether on C-10, affording 15 as a white powder (68% overall isolated yield over 2 steps, multigram-scale). |
63c835ea5ab3135f8eac3f4a | 6 | Arrival at macrocycle 15 (the end of the "cyclase phase") was a milestone since all requisite core C-C bonds were in place to arrive at 1 and 2. All that remained was installation of five oxygen atoms at C-5, C-7, C-13, C-14, and C-15. The "oxidase-phase" commenced with oxidation of C-14/15 since olefinoxidations are well-known to be compatible with alkynes. Initial attempts using oxidants (i.e., OsO4, m-CPBA, etc.) delivered undesired stereochemical outcome at C-15 in all cases (Fig. ). Therefore, a net six-electron oxidation catalyzed by ruthenium was chosen to arrive at a diketone , with an eventual strategic reduction to set the desired stereochemistry. Stronger oxidants of this type, however, will not tolerate the presence of an alkyne. For this purpose, an internal protection strategy was designed and achieved upon skeletal reorganization by refluxing 15 and XPhosAuNTf2 (0.8 mol %) in CH2Cl2. The newly-formed tricyclic system reorganized all potentially sensitive sites (nitrogen on C-1, C-4 ketone, C-7-C-8 alkyne, and C-10 alcohol) into their inactive states. Subsequent six-electron oxidation under Rucatalysis led to diketone 16 in a 53% isolated yield (gram-scale) thereby minimizing reliance on PGs. |
63c835ea5ab3135f8eac3f4a | 7 | With this newly constructed rigid polycyclic system in place, the correct oxidation state and stereochemistry on C-14 and C-15 were installed. Site-specific reduction of C-14 was accomplished using L-selectride, followed by treating the crude material with NaBH4 to afford diol 32 (see the SI), which possessed the desired stereochemistry on C-15. Subsequently, C-14 was selectively returned to the ketone oxidation state with TEMPO/NaOCl. Other oxidants tested for this step showed poor selectivity (Table ). Remarkably, upon heating crude 17 with zinc powder in acetic acid, the Troc group was dismantled and the polycyclic ring system spontaneously unraveled through ring-chain tautomerization (presumably via 18) to liberate the deoxyportimine triflate 19 in 74% yield on gram-scale. |
63c835ea5ab3135f8eac3f4a | 8 | Only two oxygen atoms, at C-13 and C-5, remained to be installed to complete the synthesis. Both oxidations could be achieved in a single step by treating the crude silyl enol ether of 19 with DMDO to afford a nitrone 20 (73% overall yield). Subsequent heating of 20 in the presence of Ac2O and TEA presumably triggers a Boekelheide type rearrangement to deliver a diacetate (compound 39, see the SI) as a single diastereomer, followed by regioselective hydrolysis of the C-5 acetate using LiOH to yield monoacetate 21 (64%) on 330-mg scale. At this point, the vinyl triflate which had remained a silent observer throughout the synthesis was now called upon to append the final two carbon atoms of 1 and 2. Thus, a Suzuki coupling was chosen to install the exocyclic vinyl group, delivering diene 22 in 75% yield. To complete the synthesis, 22 was oxidized with DMP, followed by hydrolysis, affording portimine B (PB, 2) in 88% yield. During these studies we suspected that the originally assigned structure of 2 as a ring-opened tautomer was incorrect and this was now confirmed to be the same ring-closed tautomer expressed in 1. To complete the synthesis of portimine A (PA, 1), crude 2 in methanol could be stereoselectively reduced by NaBH3CN in high isolated yield (80%). |
63c835ea5ab3135f8eac3f4a | 9 | With this scalable synthesis in hand, we next turned to determining the mechanism of action (MoA) of PA . First, we sought to identify sites that could accommodate a "fully functionalized" retrieval tag to facilitate chemical proteomic target identification while not interfering with the biological activity of the parent structure . Here, an evaluation of portimine A analogs with modifications at multiple sites (Fig. ) in Jurkat (human T lymphocyte) and HCC1806 (human breast cancer) cells confirmed that fully-synthetic PB (2) showed significantly less toxicity in both cancer cell lines compared to PA (1) (7), as did epi-portimine A (ePA, 38), the C-5 epimer of PA . Taken together, these results suggests that the anti-proliferative activity of PA (1) is dependent on the stereochemistry of C-5, and that ePA could be employed as an inactive control compound for further studies. We also noted that phenyl-derivative Ph-PA showed similar activity to PA (1) in both cell lines, indicating the terminal vinyl group (C-21/22) is not essential for the observed biological activity. Based on these results, we pursued the assembly of a diazirine-alkyne (DA) containing photoaffinity tag at the C-18 position of PA (1) through Suzuki coupling with the key intermediate (compound 21, for detailed synthesis steps, see SI). Encouragingly, we observed both the epimeric photoaffinity probes PA-DA (35-2) and ePA-DA (35-1) retained identical activities as their parent analogs, suggesting that they could serve as target ID tools (Fig. ). |
63c835ea5ab3135f8eac3f4a | 10 | It had previously been reported that isolated PA (1) induces cellular apoptosis . Indeed, we observe that fully synthetic PA (1) also induces apoptosis, as does Ph-DA and PA-DA , at low nanomolar concentrations, as determined by caspase-3 activation in Jurkat cells, however controls ePA and ePA-DA (35-1) do not (Fig. and Fig. ). Further investigation also revealed that PA (1) and PA-DA (35-2) markedly increased the proportion of Jurkat cells in G1 phases (Fig. and Fig. ), but not in ePA and ePA-DA (35-1) treated cells. Notably, we observe that PA (1) and Ph-PA display minimal effects on cell viability in freshly isolated human peripheral blood mononuclear cells (PBMCs) as well as no obvious effects on caspase activation (Fig. and Fig. ), suggesting that their toxicity mechanisms are selective to rapidly proliferating cells. |
63c835ea5ab3135f8eac3f4a | 11 | We next pursued identification of the protein targets of PA (1) in cells by tandem mass tags (TMT)-based proteomics . Specifically, we aimed to identify proteins that were substantially enriched (>2fold) by PA-DA over ePA-DA and competed (>5-fold) by PA (1) but not the inactive epimer ePA (Fig. and Fig. , Table -S7) in both Jurkat and HCC1806 cells. We identified only one protein, 60s ribosomal export protein NMD3 in both cell lines that fulfilled these criteria. We verified this interaction by chemoprecipitation (ChP), where the labeling of endogenous NMD3 by PA-DA (35-2) could be blocked by excess PA (1), but not excess ePA , and little to no labeling was observed in the absence of UV irradiation or with ePA (38) (Fig. and Fig. ), suggesting that PA (1) selectively and non-covalently binds to NMD3 in cells. NMD3 is an adaptor of 60S ribosomal subunit nuclear export and is released upon pre-60S maturation in the cytosol and subsequent polysome formation, though the precise mechanism of this regulation is not fully elucidated . Notably, we observed that NMD3deficient Jurkat and HeLa cells were less prone to the cytotoxic effects of PA (1), compared to control cells (Fig. , Fig. ), suggesting that NMD3 is necessary for its activity. As NMD3 is required for nascent 60S ribosome maturation (41), we next examined whether PA (1) affects ribosome assembly. Here, we observed that in Jurkat cells treated with PA (1), but not inactive ePA , led the accumulation of 80S monosomes and disomes as well as the blockade of polysome formation (Fig. , Fig. ). Further, we observed that eukaryotic translation initiation factor 6 (eIF6), an essential factor of 60S maturation and 80S assembly which is reported to be modulated by NMD3 , is localized in the 60S fraction and decreased in RNA-free fractions in cells treated with PA (1) (Fig. , Fig. ). Together, these data suggest that PA (1) engages NMD3, which results in increased 60S associated eIF6, stabilization of 80S, and subsequent impairment of polysome formation (Fig. ), likely leading to cell cycle arrest and apoptosis . |
63c835ea5ab3135f8eac3f4a | 12 | The scalable total synthesis of portimine A (1) and B (2) presented herein benefits from a strategy that is distinct from both biosynthesis and prior approaches. Driven by a desire to avoid an abundance of functional group manipulations, non-strategic redox fluctuations, and PGs, a plan was forged to forego the installation of oxygenation until a late stage. By analogy to two-phase terpenoid synthesis, this required the construction of a minimally decorated carbon skeleton followed by sequential oxidations. In this way, the innate reactivity of a macrocycle could be leveraged to install both the correct oxygenation pattern as well as the stereochemistry. Certain intermediates served to "self-protect" key functional groups and enable the final sequence by strategically timed ring-chain tautomerization events. Finally, an unusually stable vinyl triflate was carried through the majority of the synthesis allowing for flexible diversification of this biologically promising lead compound, enabling the construction of photoaffinity probes for target identification studies. With sizable quantities of PA (1) in hand, we have shown that PA (1) potently induces G1 cell cycle arrest and apoptosis in multiple human cancer cell lines, but not in human PBMCs. Further, chemical proteomic studies revealed NMD3 to be the primary target of PA , and to our knowledge, is the first reported small molecule ligand of NMD3. Engagement of NMD3 by PA results in a buildup of 80S ribosome and blockade of polysome formation. Further, the observation that decreasing NMD3 expression reduces PA potency is suggestive of a gain-of-function or neo-function mechanism, however, the molecular details of how binding of PA (1) to NMD3 induces this cascade are not yet clear and will be subject of future studies. Overall, this study demonstrates the utility of totally synthetic routes designed with ideality criterion combined with powerful chemical proteomic methods to reveal new potential therapeutic targets . |
65ca149e66c138172975f523 | 0 | We rationalize the success of these adjustments in terms of the specific physical-chemical properties of TMDCs, namely their anisotropic in-plane/out-of-plane carrier behavior, large optical absorption, and chalcogenide-dependent surface chemistry. Just one surprisingly simple yet effective pathway to fast TMDC photodetection is the reduction of the photoresistance by using light-focusing optics, which enables bandwidths of 0.23 GHz with an energy consumption of only 27 fJ/bit. By reflecting on the ultrafast intrinsic photoresponse times of few picoseconds in TMDC heterostructures, we motivate the application of more demanding chemical strategies to exploit such ultrafast intrinsic properties for true GHz-operation in real devices. A key aspect in this regard is the management of surface defects, which we discuss in terms of its dependence on the layer thickness, its tunability by molecular adlayers and the prospects of replacing thermally evaporated metal contacts by laser-printed electrodes fabricated with inks of metalloid clusters. We highlight the benefits of combining TMDCs with graphene to heterostructures that exhibit the ultrafast photoresponse and large spectral range of Dirac materials with the low dark current and high responsivities of semiconductors. We introduce the bulk photovoltaic effect in TMDC-based materials with broken inversion symmetry as well as a combination of TMDCs with plasmonic nanostructures as means for increasing the bandwidth and responsivity simultaneously. Finally, we describe the prospects of embedding TMDC photodetectors into optical cavities with the objective of tuning the lifetime of the photoexcited state and increasing the carrier mobility in the photoactive layer. |
65ca149e66c138172975f523 | 1 | Transition metal dichalcogenides (TMDCs), especially two-dimensional monolayers, exhibit strong excitonic binding energies (300 meV), large carrier mobilities (30 cm 2 /Vs) and high extinction coefficients (~3*10 6 cm -1 ). They are chemically versatile, compatible with silicon technology, relatively cheap, and can be used as monolayers, multilayers or combined with other 2D materials in heterostructures. These properties are ideal for their application in photodetectors with high speed and large detectivity at the same time. However, as detailed by Sorger et al. an independent optimization of speed and detectivity (or responsivity) is not possible, since the long carrier lifetimes needed for high responsivities are detrimental for the response time. Similarly, large mobilities are required for fast photodetection with sufficiently large band gaps to prevent high dark currents and, thus, low detectivities as e.g. in zero-gap graphene. Over the past ten years, most efforts have focused with impressive results on increasing the responsivity. In contrast, similar improvements of the response speed of TMDC photodetectors have been challenging, and in particular a GHz photoresponse has remained mostly elusive, despite an increasing demand for GHz-compatible photodetectors as components in optical data communication. In this account, we reflect on the lessons learnt from our own physical-chemical approaches toward ultrafast TMDCs photodetectors. |
65ca149e66c138172975f523 | 2 | The article is structured as follows: We begin with a brief account of the fundamental performance parameters of photodetectors for readers new to the field (see Box 1). We continue with a definition and distinction of extrinsic vs. intrinsic response times, discuss the effect of the TMDC layer thickness and provide a brief overview of the speed limiting factors in TMDC photodetectors. On this basis, we devise chemical and physical design strategies for accelerating their photoresponse, taking into account substrate considerations, chemical healing of defects, edge electrodes and reducing the photoresistance. We conclude with promising future directions to even faster photodetection, including optical cavities, the bulk photovoltaic effect and plasmonics. |
65ca149e66c138172975f523 | 3 | Box 1: Figures of merit of photodetectors Selection of sensitivity measures: The time it takes the current to increase from 10 to 90% (90 to 10%) of its final value is called the rise (fall) time. Two different laser excitations are distinguished: Steady (a) and non-steady state (b) excitation, as sketched below. In the steady state a square pulse illumination increases the current to the highest possible photocurrent. The non-steady state mimics real data transfer by illumination with a delta shaped impulse and observing the response time. Via the power spectrum, the frequency-based speed measure, the 3dB bandwidth, can be determined, giving the maximal operation frequency at which the initial photocurrent has dropped to 70%. Note that the often-used approximation 𝑡 𝑟𝑖𝑠𝑒 = 0.35 𝑓 3𝑑𝐵 ⁄ can lead to large deviations compared to the much more precise power spectrum, when calculating the bandwidth. |
65ca149e66c138172975f523 | 4 | The extrinsic photoresponse is the true signal processing time (detector speed) of the entire device with highest relevance for real applications. The basic concepts for its experimental determination are detailed in Box 1. A laser with a pulse length much shorter than the response time excites the photoactive material to assess the rise and fall times, limited by the slowest mechanism present. Possible limitations are the RC time, drift / transit time and diffusion time. The RC time 𝑡 𝑅𝐶 is the product of the device capacitance and the device resistance, which is omnipresent in every electrical component. included in these fundamental three components are the carrier injection time at the electrode and the transfer time between two materials in a heterostructure or even between two layers in a multilayer. The injection time or electrode-material transfer time is found indirectly in the RC time in form of the contact resistance whereas the material-transfer time is part of the drift or diffusion time, depending on the depletion. |
65ca149e66c138172975f523 | 5 | Its determination requires all-optical methods, such as the two-pulse coincidence (2PC) technique, which is detailed in Box 2. Briefly, a pulsed pump laser excites the sample and after a short delay, a second laser, the probe laser, examines how many charge carriers can be re-excited. The temporal resolution of this experiment is given by the pulse width of the pump laser and/or the delay time between the two lasers, such that measurements with dt < 100 fs are possible. |
65ca149e66c138172975f523 | 6 | τin -1 = (τd + τs) The working principle of asynchronous optical sampling (ASOPS) is similar, however, a detuning repetition frequency Δf of the pump laser compared to the probe laser is adopted to form a pump-probe delay. With the repetition frequency difference Δf, when the first pulses of the pump laser and the probe laser coincide, a successive offset for the subsequent pulses as a delay time will be generated. Delay times range from femtoseconds to several ns (5 ns for 100 MHz repetition rate). 4 a-d) Adapted with permission from ref. Copyright 2022 American Chemical Society. |
65ca149e66c138172975f523 | 7 | A key feature of most TMDCs is the direct-to-indirect bandgap transition upon crystal growth from a truly two-dimensional monolayer to multilayers. Since direct optical transitions exhibit larger extinction coefficients and usually faster recombination times, it is generally assumed that TMDC monolayers are superior for photodetection in both, sensitivity and speed. Indeed, for the intrinsic photoresponse this holds true, as demonstrated e.g. for two terminal MoTe2 photodetectors with thicknesses of 2 nm to 35 nm and response times of 4 ps to almost 1 ns, respectively (Figure ). In addition to the effect of the direct-to-indirect bandgap transition, this finding could be rationalized further in terms of the quantum mechanical wave function model, predicting that bulk defects in MoTe2 have a much longer lifetimes than surface defects, which increases the intrinsic response time. For the extrinsic photoresponse however, additional factors must be considered, which may outcompete the intrinsic advantages of monolayers. These include the stronger total absorbance of multilayers (due to longer optical path lengths) and the concomitant increase in photocurrent as well as material-specific differences in the lifetimes of surface defects. For MoS2 for instance, monolayers exhibit three orders of magnitude longer extrinsic response times than multilayers of 10 nm thickness, presumably due to deep surface trap states, which become increasingly screened in the bulk. This behavior is also often referred to as persistent photocurrent, a feature that is quite prominent in MoS2 and severely limits the prospects of MoS2 monolayers as fast photodetectors. For WSe2 in contrast, we have recently shown that the extrinsic response time is much more robust against surface trap states. Here, the major speed limitation is the RC time, which scales with the photoresistance of the devices. Monolayers of TMDCs exhibit less total absorbance and, thus, fewer photoexcited charge carriers which leads to longer RC times. In principle, this disadvantage can be addressed by increasing the irradiance as we will detail below. |
65ca149e66c138172975f523 | 8 | The results and conclusions in this section are based on our work with two-terminal lateral photodetectors based on MoS2 or WSe2, which we use as a model device architecture due to its cost-effective fabrication. A key result of these studies is that such simple photodetectors are essentially all limited by the RC constant of the device. While this may not necessarily be true for more complicated TMDC photodetectors, such as gated three-terminal devices, vertical geometries or heterostructures, we believe that the considerations here should be widely applicable also to other TMDC materials within two-terminal lateral detectors. Therefore, this section focuses on strategies for decreasing the RC time, which are validated by monitoring the overall effect on the extrinsic photoresponse. |
65ca149e66c138172975f523 | 9 | To illustrate the importance of the substrate material on which the TMDC is deposited, we focus on the surface roughness and the dielectric constant. The substrate roughness is particularly relevant for TMDC monolayers as an uneven surface enhances carrier scattering. Moreover, since most carrier transport occurs in the few TMDC layers located closest to the electrodes, such carrier scattering also affects bulk TMDC crystals. We have measured bulk WSe2 on (smooth) glass and |
65ca149e66c138172975f523 | 10 | (rough) polyimide substrates. Polyimide yields lower dark currents by more than two orders of magnitude, cf. Figure , presumably due to its rougher surface (30.7 ± 11.6 nm vs 5.5 ± 0.7 nm on glass). Such an increase in the resistance is detrimental to the RC time and the expected response time. In a similar context, Haizmann et al. have recently shown that substrate roughness and its effect on thin TMDC layers also alters the orientation of small molecules adsorbed to the TMDC surface. While the exact consequences of this alteration remain to be explored, it is likely that this will affect the chemical interactions at the interface and, thus, the efficiency with which such small molecules saturate surface defects. Therefore, a popular and powerful strategy is the insertion of an atomically smooth protection layer of hexagonal boron nitride underneath the TMDC layer. This action improves the charge carrier mobility, however at the expense of additional fabrication steps. from Ref. with permission from the PCCP Owner Societies. |
65ca149e66c138172975f523 | 11 | For bulk WSe2, we found that the dielectric constant of the substrate not only alters the response time but also the speed-limiting mechanism. Glass has a larger dielectric constant of 4.5 -8, in comparison to the weaker screening on polyimide with 3.7. For polyimide, the WSe2 detectors are faster and limited by the RC time. For the detectors on glass, the response time is slower than the RC constant, presumably now limited due to drift limitation as a consequence of a shrunken depletion layer induced by the higher dielectric screening (Figure ). The maximal achieved bandwidth is 2.6 MHz. |
65ca149e66c138172975f523 | 12 | Layered materials offer several ways of contacting: bottom electrodes beneath the TMDC, top electrodes above the material, and a sandwich structure in a vertical stack or so-called "edge contacts" at the edges of the material. A key result of our work with MoS2 is that edge contacts provide unique advantages for the speed of photodetectors based on anisotropic 2D materials, such as TMDCs. With the unravelling of the edge, for example by plasma etching, there is direct access to the layers which differ chemically tremendously from the surface. They have been shown to provide a lower contact resistance, a smaller transfer length or a higher capability of charge carrier injection. We find that this contact style accelerates the fall time of the photodetectors by more than an order of magnitude (Figure ). This result is presumably the combination of three effects: |
65ca149e66c138172975f523 | 13 | First and foremost, by contacting the edge of the TMDC, access to the much higher in-plane mobility is given. No van-der-Waals gaps between layers have to be overcome, instead the excitons in the layers can be extracted in parallel and the interlayer transfer time is irrelevant for the photoresponse. Second, the work function of the MoS2 on the edges is different than on the top facet, which affects the magnitude of a potential Schottky barrier at the TMDC/metal interface. Both of these effects take specific advantage of the anisotropic in-plane/out-of-plane carrier behavior in TMDCs. A third effect arises from the fact that top contacts often use adhesion layers, e.g. titanium, between the TMDC and the actual metal contact, e.g. gold. In the edge contact geometry, there is almost no contact between the TMDC and the adhesion layer, which again affects the height of a potential Schottky barrier at the interface. For the specific example of titanium vs. gold, we find that pure titanium edge contacts are faster than the top devices, but slower than Au edge devices. This is in agreement with Zhang et al. who correlated the faster response times for Au contacts with a higher Schottky barrier. All advantages of edge contacts combined lead to accelerated rise and fall times, also visible when comparing the non-steady state and Fourier transform. The representative example shown in With respect to the widely applied top contact architectures in the design of TMDC photodetectors, we note an additional complication: these contacts are typically fabricated by thermal evaporation of the metal in vacuum. Hot metal atoms induce defects and alloy formation in the top layers of the TMDC, leading to Fermi-level pinning and other potentially unwanted changes to the surface with detrimental effects on the speed of the photodetector. Several strategies have been designed to circumvent this complication, including transfer printing of metal electrodes via polymer stamps or so-called van-der-Waals electrodes. Geladari et al. have recently shown that solutions of atom precise Au32-metalloid gold clusters can be utilized as inks for direct laser-induced printing of gold contacts with diffraction-limited spatial resolution. The clusters are separated and stabilized with (nBuP12Cl8) linkers. Upon illumination at 488 nm these linkers are detached from the Au cores, which then fully agglomerate to macroscopic, metallic gold. This way, gold contacts with near bulk-like conductivities (10 6 S/m) were defined on WS2, and a fully functional photodetector was fabricated without thermal evaporation, cf. Figure . The thermal stress inflicted by such methods is only given by the absorption of light, which strongly limits the imposition on the WS2 flake. While quantitative details remain to be fully investigated, we expect such gentle contacting strategies to be advantageous in reducing the carrier injection time, which would manifest in a reduced RC time. Ref. Copyright 2023, Wiley-VCH. |
65ca149e66c138172975f523 | 14 | An important observation from our work with TMDC photodetectors is that under typical optical excitation conditions (irradiance between 0.4 W/cm 2 and 400 W/cm 2 ), the photoactive material does not saturate. This means that larger irradiances -either due to higher laser power or better optical focusing -invoke lower photoresistance since additional charge carriers are excited. As the RC constant of a photodetector scales directly with the photoresistance (not the resistance in the dark), increasing the irradiance is a simple yet effective means for increasing the speed of TMDC photodetectors. This becomes apparent in Figure (a) which depicts the dependence of the response time in WSe2 multilayers on the photoresistance, which is varied solely by changing the irradiance. The strong negative correlation indicates the aforementioned RC limitation, which is further verified by quantitative agreement with the calculated and fitted RC constants of the devices. Another effective means for decreasing the RC time is a reduction of the device dimensions, which lowers the resistance and capacitance simultaneously. By reducing the channel width from 25 to 20 µm (80 µm in the previous studies with substrate effects 3 ), the electrode width itself from 10 to 1 µm and the channel length from 2.5 to 1 µm, the capacitance is approximately reduced from 7.6 to 4 fF (24 fF previously 3 ). For this example, we obtain a response time < 2 ns and a 3dB bandwidth of 230 MHz compared to roughly 30 MHz for the larger device geometry (Figure ). |
65ca149e66c138172975f523 | 15 | We note that this bandwidth is a conservative, lower-bound estimate, since such WSe2 are so fast that they exceed the current speed limit of our set-up as verified with a commercial photodiode In the context of energy-efficient optical communication, one may be concerned with the energy consumption of such a photodetector considering that the irradiance required for optimal speed of roughly 100 W/cm 2 is quite high. For our example and a bandwidth of (at least) 230 MHz, a laser power of 6.25 µW is sufficient. Since the detector is operated at zero bias by taking advantage of the photovoltaic voltage generated at the metal/semiconductor junction, the laser power is the only significant contribution to the overall energy consumption, which amounts to < 27 fJ/bit. This compares favorably to most optical communication components (>1000 fJ/bit) and is not much higher than the energy required to switch a simple CMOS gate (~1 fJ/bit). Exploring and expanding the ultimate speed limit for TMDC photodetectors |
65ca149e66c138172975f523 | 16 | With the strategies detailed above in place, one may expect to finally lift the current RC limitation of two-terminal TMDC photodetectors. Ultimately, and under the assumption that drift as well as diffusion times remain insignificant, the speed limit of such optimized devices will be determined by their intrinsic material response times. Assessing and optimizing these is therefore relevant in determining materials and architectures that provide the highest prospects for true GHz performance in optical applications. For example, 2PC measurements of an MoS2 monolayer in a two-terminal photodetector with a simple metal-semiconductor-metal configuration without external bias has revealed an intrinsic response with a fast component of 3-5 ps and a slow component of 80-100 ps. This picosecond response is derived from defect-related recombination processes. Upon photoexcitation, the thermalization and cooling of excited carriers occurs on the time scale of 0.5-1 ps. After that, these charge carriers are captured by different defects such as surface defects and grain boundaries with different lifetimes, which determine the intrinsic speed limit of the photodetector. |
65ca149e66c138172975f523 | 17 | 2PC measurements of the intrinsic response time may also be used to explore the potential of more complicated TMDC device structures, such as vertical van der Waals p-n junction heterostructures. The strong built-in electric field in these junctions has been shown to greatly reduce the carrier drift time. This mitigates the detrimental effect of thick multilayers (see the example of MoTe2 in Figure ) and allows exploiting the larger total absorbance of bulk crystals. |
65ca149e66c138172975f523 | 18 | An example is the MoS2/25 nm WSe2 heterostructure depicted in Figure Heterostructures with ultrafast intrinsic response times may also be used to design TMDC-based photodetectors for operation in the infrared region. With pure TMDCs, this is challenging due to their large bandgaps, especially when targeting the important telecommunication band at 1560 nm. |
65ca149e66c138172975f523 | 19 | This restriction may be lifted by integrating graphene into the TMDC photodetector. In this configuration, the low-energy photon-induced hot carriers from graphene can be injected into the TMDC and the spectral window of the TMDC photodetector is broadened. The generated charge 20 carriers in graphene can relax via carrier-carrier scattering without the assistance of the lattice. For pure graphene, this advantage becomes apparent in terms of an intrinsic response time of 1.5 ps at room temperature. However, pure graphene exhibits high dark currents and, thus, low detectivities. Combining graphene with WS2 to a van-der-Waals heterostructure preserves the ultrafast hot carrier cooling and results in a response time of 1.2 ps under 1560 nm excitation (Figure ). This photodetector combines the advantage of the low dark current in TMDCs and the fast hot carrier cooling as well as infrared sensitivity of graphene. If the extrinsic limits of these devices could be lifted at least to some extent, these ultrafast intrinsic response times suggest that GHz device operation should be easily feasible with TMDCs, even in the infrared regime. |
65ca149e66c138172975f523 | 20 | Implementing a photodetector into an optical cavity, e.g. a Fabry-Pérot resonator, leads to a confinement of the electromagnetic field and an alteration of the emission rate of the photoactive material. By tuning the length of the cavity, optical transitions in resonance with a cavity mode are selectively amplified (Figure ). This coupling can have several advantageous effects for fast photodetection, such as the tuning of the lifetime of the photoexcited state or an increased carrier mobility. The latter has been demonstrated for a perylene diimide derivative and resulted in an increase of one order of magnitude, presumably due to a polariton-mediated delocalization of excitons. For the coupling to be strong, it is important that the reabsorption of emitted photons is efficiently possible, which favors materials with small Stokes shifts, such as WS2. One can expect that strong coupling between photodetector and optical cavity will drastically enhance the speed and efficiency of the device, which is the basis for the emerging field of polaritonic chemistry. 39 |
65ca149e66c138172975f523 | 21 | Conventional photodetectors are subject to a trade-off between speed and responsivity. In photoconductive devices it is desirable for the responsivity to create gain via long lifetimes of the minority carriers, but this action simultaneously invokes long response times. Here, the so-called bulk photovoltaic effect (BPVE) can be advantageous, which is a nonlinear optical process that occurs in non-centrosymmetric materials, creating an anomalous current. spontaneous electronic polarization to separate charge carriers to result in theoretically high photoelectric conversion efficiencies. Potential candidates for its exploitation are the rhombohedral 3R-phase of many TMDCs that naturally possess an out-of-plane polarization (Figure ) or heterostructures where the required breaking of the inversion symmetry is created at the interface. In a graphene/3R-MoS2/graphene heterostructure, this concept demonstrated an intrinsic response time of 2 ps, indicating a high photodetection bandwidth comparable to graphene. Moreover, the conversion efficiency for the BPVE is related to the free path length (l0), which refers to the non-thermalized carrier transit length before they descend to the bottom of the conductive band. Due to the small thickness of the graphene/3R-MoS2/graphene heterostructure, the vertical charge carrier collection distance is smaller than l0, resulting in a responsivity of 70 mA/W -1 and EQE of 16 %, which is large considering the fast speed of the detector. 42 |
65ca149e66c138172975f523 | 22 | Another approach to simultaneously enhance the speed and responsivity of photodetectors is the coupling to localized surface plasmon resonances (LSPR). A localized surface plasmon is the collective oscillation of the free electron gas within a metallic nanostructure, which decays on the timescale of picoseconds after excitation. The radiative component of this decay plays a key role in the near-field enhancement of the electric field, while the non-radiative component generates hot electrons (Figure ). The resonances, decay components and -times can be tuned by the chemical nature and morphology of the nanostructure, e.g. with sharp structures like stars. A fast collection of the hot electrons provided, plasmonic/TMDC detectors exhibit bandwidths in the GHz regime. The simultaneous near-field enhancement of plasmonic structures is advantageous for the responsivity, and an 1000-fold increases in photocurrent as well as the extension of the spectral range has been demonstrated for a Pt-plasmonic/MoS2 photodetector. The combination of ultrafast hot electrons, high near field enhancements, potential use of plasmonic lattices and the extended detection range hold great promise to reach tens of GHz bandwidths with good responsivities in TMDCs. 50 |
65ca149e66c138172975f523 | 23 | If these trap states are located deep within the band gap (as for instance in MoS2), the resulting long lifetimes of the trapped carriers greatly decrease the speed of the photodetector. A promising strategy to mitigate this inherent disadvantage of 2D-materials is the formation of van-der-Waals heterostructures employing molecular adlayers, such as metal phthalocyanines. Haizmann et al. |
65ca149e66c138172975f523 | 24 | have shown by angle-resolved photoelectron spectroscopy that the adsorption of perfluorinated cobalt-phthalocyanine (CoPcF16) on bulk n-type MoS2 restores an intrinsic position of the Fermi level, indicating that trap states near the conduction band edge could be compensated by the adlayer. This action was accompanied by the appearance of a new gap state near the Fermi level, which was strongly confined to the interface (Figure ). In contrast, the fluorine-free analogue CoPc did not induce a new gap state but increased the degree of n-doping in the heterostructure. |
62b03bf3eb1f0cecbb8f8f73 | 0 | Interest in the magnetic properties of polymetallic clusters of Ni II began with the development of magneto-structural correlations of [Ni2] dimers 1 and [Ni4] cubanes 2 that revealed a dependence of the sign and magnitude of the exchange interaction on both the Ni-X-Ni bridging angle and the anisotropy of the Ni II ion. The large axial zero-field splitting displayed by the latter in certain geometries also lends itself to the construction of both single-molecule magnets (SMMs) and single-ion magnets (SIMs) displaying slow relaxation of the magnetisation. Indeed, recent studies of Ni II SIMs at both ambient and high pressure have revealed how magnetic anisotropy is extremely susceptible to even small structural distortions, in turn highlighting target geometries and directing the synthetic methodologies required to engineer molecules possessing giant magneto-anisotropies. Flexible N,O-bridging ligands have proved particularly successful in the construction of polymetallic clusters of Ni II displaying a variety of topologies and nuclearities, including supertetrahedra, 9 wheels, 10 planar discs and icosahedra. The pro-ligand (3,5-dimethyl-1H-pyrazol-1-yl)methanol (HL 1 ) belongs in this family, having been employed to make both mono-and tetranuclear clusters of Ni II . Here, we expand this chemistry to include the synthesis, structure and characterisation of [Ni14(HL 2 )12(HCOO)14Cl14(MeOH)(H2O)]•4Me2CO (1•4Me2CO, HL 2 = 3,5-dimethylpyrazole), an aesthetically pleasing wheel formed serendipitously via the in-situ transformation of HL 1 to HL . |
62b03bf3eb1f0cecbb8f8f73 | 1 | The reaction of NiCl2•6H2O and HL 1 in a basic MeOH solution heated at 65 °C for 40 minutes affords compound 1 (Fig. ) upon diffusion of acetone into the cooled mother liquor (see the SI for full experimental details). Crystals of 1 are in a tetragonal crystal system and structure solution was performed in the space group P42/n (see the SI for full crystallographic details, Table , Fig. ). The asymmetric unit contains half the formula unit. |
62b03bf3eb1f0cecbb8f8f73 | 2 | The metallic skeleton of 1 is a single stranded wheel (Fig. ). The bridging between each pair of Ni II ions is the same around the entire wheel and consists of one µ-Cl ion (Ni-Cl-Ni, ~82.0-85.1°), and one µ-O atom (Ni-O-Ni, ~102.8-103.2°) and one µ-carboxylate which both derive from the syn, syn, anti-bridging formate (Fig. ). The six-coordinate Ni II ions are all in distorted octahedral geometries with their {NiO3Cl2N} coordination spheres completed by a terminally bonded HL 2 ligand. The latter and the formate ions originate from the in-situ reaction of HL The only exception to this is Ni5, {NiO4Cl2}, in which there resides a disordered MeOH/H2O molecule in place of the HL 2 ligand. The wheel is non-planar with nearest neighbour Ni II ions being above and below the plane running through the middle of the fourteen metal ions, i.e., they form a zigzag/sinusoidal "up-down-up-down" motif as the wheel is circumnavigated (Fig. ). The approximate dimensions of the wheel are, Ni1 Dc magnetic susceptibility (χ) and magnetisation (M) measurements of 1 were taken in the T = 300-1.80 K, B = 0.1 T and T = 2.0-10 K and B = 0.5-9.0 T temperature and field ranges, respectively. These are plotted as the χT product versus T, and M versus B in Fig. . The T = 300 K value of χT = 18.5 cm 3 K mol -1 is equal to the value expected for fourteen non-interacting Ni II ions with g = 2.30. Upon cooling the χT value remains relatively constant, increasing only very slowly to ~24 cm 3 K mol -1 at 50 K before rising sharply to a maximum of ~104 cm 3 K mol -1 at T = 3 K. The value then drops to ~96.3 cm 3 K mol -1 at 2 K. The M vs B data increases rapidly with increasing field, saturating at a value of M = 32.1 µB at T = 2 K and B = 9 T. The susceptibility and magnetisation data are therefore indicative of weak ferromagnetic nearest neighbour exchange and the stabilisation of an S = 14 spin ground state. |
62b03bf3eb1f0cecbb8f8f73 | 3 | The magnetic susceptibility data can be simulated using exact diagonalisation 18 and an isotropic spin-Hamiltonian 𝐻 ̂= -2 ∑ 𝐽 𝑖𝑗 𝑖<𝑗 𝑠 ⃗ ̂𝑖 • 𝑠 ⃗ ̂𝑗 with a coupling scheme that assumes just one independent exchange interaction between nearest neighbours, J = + 4 cm -1 with g = 2.30 (Fig. , black curve). The addition of a next nearest neighbour interaction makes no difference to the quality of the simulation. Given that this interaction is computed to be very weak and ferromagnetic by DFT (vide infra) this is to be expected. The DFT calculated values for the seven crystallographically unique interactions also simulate the susceptibility well if they are scaled by a factor of 1.4 (Fig. , red curve). This simple isotropic model, however, does not explain the low temperature magnetisation data, which requires inclusion of the single ion anisotropy of the Ni ions, D(Ni), to be included 𝐻 ̂= 𝐷 ∑ (𝑠 ⃗ ̂𝑖 • 𝑒 ⃗ 𝑖 ) |
62b03bf3eb1f0cecbb8f8f73 | 4 | , where 𝑒 ⃗ 𝑖 = 𝑒 ⃗ 𝑖 (𝜗 𝑖 , 𝜑 𝑖 ) is the direction of the local easy axis. Computational limitations direct us toward employing a [Ni7] wheel, with the results multiplied by two to mimic the [Ni14] wheel. The magnetisation data is simulated nicely with J = +4 cm -1 and D(Ni) = -5 cm -1 with the anisotropy axes tilted from the axis of the wheel by θ = 30°, 𝜑 𝑖 = 2𝜋𝑖/7, in agreement with ab initio NEVPT2 calculations (see Figure , blue curves, and below). The magenta curves in Fig. demonstrate for the isotropic case that the substituted [Ni7] model system is close to the original except for low temperatures where the S = 14 ground state cannot be reproduced. The ferromagnetic exchange in 1 is consistent with magneto-structural correlations developed for halide-bridged Ni II dimers where the sign and magnitude of the interaction is dictated by the Ni-Cl-Ni angle -with a switch from antiferromagnetic to ferromagnetic occurring at approximately ≤102° and increasing with decreasing angle. Note the Ni-Cl-Ni angles in 1 are ~82-85°. To further understand the origin and sign of the magnetic coupling constants we have performed DFT calculations on models created from 1 (1A-C, Fig. , Tables ). All seven unique exchange interactions are in the range +1.7 ≤ J ≤ +3.9 cm -1 , consistent with the experimental values. The narrow range of values obtained can be attributed to the presence of similar structural parameters for each metal ion, with the relatively small Ni-µ-O/Cl-Ni angles resulting in ferromagnetic exchange (Table ). To further explore the origin of the sign and magnitude of these interactions we have performed overlap integral calculations between the singly occupied molecular orbitals (SOMOs) of the Ni II ions in a bimetallic model (1D) created from 1 (Fig. ). These calculations suggest competition between one moderate interaction [<Ni(α)dx 2 -y 2 ||Ni(β)dz 2 >] and three weak interactions [<Ni(α)dx 2 - y 2 ||Ni(β)dx 2 -y 2 >; <Ni(α)dz 2 ||Ni(β)dx 2 -y 2 >, <Ni(α)dz 2 ||Ni(β)dz 2 >]. The former contributes to the antiferromagnetic and the latter to the ferromagnetic part of the exchange. In this case, the three weak interactions dominate and the overall result is the observation of a weak ferromagnetic interaction. Spin density analysis suggests a strong spin delocalisation mechanism, with the spin densities on the Ni II ions being between 1.668-1.682. Of the three different bridging moieties, the Cl ion has the largest spin density (0.122 -0.141; Fig. ). To further investigate the contribution from the µ-Cl ion to the total magnetic exchange, we have replaced it with a point charge in model 1D. This results in an antiferromagnetic interaction, changing from +2.3 cm -1 to -6.2 cm -1 , clearly suggesting the major ferromagnetic contribution to the exchange comes from the µ-Cl ion. Bearing in mind the connectivity of next-nearest neighbour Ni II centres through a formate group, we have also estimated the next-nearest neighbour magnetic exchange interaction using model 1E (Fig. ). This is estimated to be very small and ferromagnetic, J = + 0.5 cm -1 . All the Ni II ions in 1 possess slightly distorted octahedral geometries (Table ) and are therefore expected to have axial zero-field splitting parameters of the order D ≤ -10 cm -1 . Ab initio NEVPT2 calculations performed using ORCA 23 reveals values of the D ≤ -6.5 cm -1 , with the major contribution arising from the dxy dx 2 -y 2 electronic transition (Fig. , Tables ). The D(Ni) axes are oriented approximately along the N(pyrazole)-Ni-O and O(MeOH)-Ni-O vectors, tilted at angles of θ = ~32-37° from the axis of the wheel. |
62b03bf3eb1f0cecbb8f8f73 | 5 | In summary, the in-situ transformation of HL 1 to HL 2 and concurrent formation of formate anions results in the self-assembly of an aesthetically pleasing [Ni14] wheel, with subsequent examination of reaction conditions leading to a more 'rational' synthetic procedure. Magnetic measurements reveal weak, ferromagnetic exchange interactions, with the susceptibility data simulated with a single exchange constant, J = + 4 cm -1 . The DFT computed values also simulate the data well, albeit they need scaled by a factor of 1.4. The simulation of the magnetisation data requires inclusion of D(Ni) = -5 cm - 1 tilted at an angle θ = 30° with respect to the axis of the wheel. Theoretical calculations are in agreement with experimental observations, revealing the major contribution to the ferromagnetic exchange is mediated through the bridging Cl ions. Attempts to make analogues of compound 1 containing different M II ions and other bridging halides, pseudohalides and carboxylates are underway. |
62e10043a8e4dcaa571ecb69 | 0 | The Nipah virus (NiV), spread by bats and can cause fatal encephalitis in people, has recently been identified in Malaysia, Bangladesh, Singapore, and India . It belongs to the order Mononegavirales, which contains other developing lethal zoonotic viruses, including Hendra, Marburg, and Ebola . The virus is thought to be stored naturally in the bodies of Pteropus fruit bats. Humans got NiV from pigs, the intermediate hosts of the virus, in 1998 during the first documented epidemic in the Malaysian town of Sungai . Since 2001, the intake of raw date palm sap contaminated with the saliva and excreta of the bats has been reported as the source of yearly NiV outbreaks in various districts of Bangladesh. The first epidemic in India was recorded in Siliguri, West Bengal, in 2001, and it was mainly spread by intimate personal contact or nosocomial transmission. In 2007, a second outbreak was reported in Nadia and West Bengal . In a recent NiV epidemic in the Kozhikode region of Kerala, a state in South India, the index patient was said to have been infected by fruit-eating bats . While nosocomial transmission accounted for the vast majority of cases, no clinical or statistical data was provided to confirm the frequency of the illness. The most recent epidemic in Kerala had a death rate of 91%, which is typical of all outbreaks . |
62e10043a8e4dcaa571ecb69 | 1 | Cell-cell fusion (syncytia) in lung, brain, kidney, and heart tissues is caused by Nipah (NiV) and Hendra (HeV) viruses. This results in encephalitis, pneumonia, and frequent death. Henipavirus infections are characterized by membrane fusion, which is required for viral entry and virusinduced cell-cell fusion . Understanding the pathobiology of henipaviruses relies on elucidating the mechanism(s) of membrane fusion, which may lead to discovering new approaches to creating antiviral therapeutics. Viral attachment (G) and fusion (F) glycoproteins must work together to facilitate membrane fusion in henipaviruses. Current theories of henipavirus fusion propose that F is released from its metastable pre-fusion conformation to promote membrane fusion after NiV or HeV G attachment to its cell surface receptors . The selected protein for this study is a fusion protein of Nipah henipavirus associated with viral infections. The physicochemical characteristics and anticipated protein structures of the selected protein demonstrated structure-function relationships of the proteins associated with viral infections. Therefore, this protein can be targeted for predicting antiviral drugs and vaccines against the selected protein to combat viral infections. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.