Pith. sign in

REVIEW 4 major objections 6 minor 126 references

RAPID-Net: Accurate Pocket Identification for Binding-Site-Agnostic Docking

T0 review · 4 major / 6 minor · reviewed 2026-08-09 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read RAPID-Net claims that precise binding-pocket identification, not ligand sampling or receptor flexibility, is the decisive factor in blind docking, and shows that a lightweight pocket predictor guiding AutoDock Vina beats DiffBindFR and…

desk verdict Solid pocket-guided docking work with a useful ranking-vs-sampling insight, but the headline benchmarks need a train/test overlap check before I'd trust the numbers. read the letter →

arxiv 2502.02371 v2 pith:NZSDBSOQ submitted 2025-02-04 q-bio.BM cs.AIcs.LGphysics.bio-phphysics.med-ph

Yaroslav Balytskyi , Inna Hubenko , Alina Balytska , Christopher V. Kelly This is my paper · ORCID
classification q-bio.BMcs.AIcs.LGphysics.bio-phphysics.med-ph
keywords Protein-ligandinteractionsBlinddockingBindingpocketpredictionSoftmasksegmentationDeepresidualnetworkConvolutionalneuralPoseBustersbenchmarkEnsemble
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that in binding-site-agnostic docking, the deciding factor is not how the ligand is sampled or how flexible the receptor is, but how precisely the binding pocket is located. To make that case, it introduces RAPID-Net, a deep residual network that predicts ligand-binding pockets as soft voxel masks and hands the resulting search box to AutoDock Vina. Guided by RAPID-Net, Vina reaches 54.9% Top-1 PoseBusters-valid poses on the PoseBusters benchmark, beating the flexible-receptor blind-docking tool DiffBindFR (49.1%), and 53.1% on the hard time split, within six points of AlphaFold 3 (59.5%) at a fraction of the cost. The authors also report that in 92.2% of cases at least one pose in the ensemble is correct, which they read as evidence that pose ranking, not sampling, is the main remaining bottleneck. If correct, the result makes accurate pocket finding the highest-leverage component of a practical blind-docking pipeline.

What carries the argument

The central object is RAPID-Net, a soft-segmentation 3D U-Net-style convolutional network that predicts a continuous per-voxel score for ligand-binding propensity rather than a binary pocket mask. Three design choices carry the argument: a ReLU activation in the final layer instead of sigmoid, a soft Dice loss based on L2 norms, and threshold-less training labels from VolSite/sc-PDB that extend beyond the 6.5 Å shell used by earlier predictors. Five independently trained replicas are combined by majority voting for high-confidence pockets and by 'minority-reported' union for recall; the resulting voxel sets define docking search grids of expandable size, which is what links pocket quality directly to downstream docking accuracy.

What would settle it

Run RAPID-Net plus AutoDock Vina on the PoseBusters complexes with each holo receptor replaced by its apo or AlphaFold-predicted structure, keeping the same pocket-to-grid protocol; if the Top-1 PoseBusters-valid rate falls far below 54.9%, the claim that pocket identification is the decisive driver would be shown to depend on seeing the bound pocket.

Watch

Extended reading notes

Core claim

The central claim is that precise pocket identification is the decisive driver of docking success in binding-site-agnostic settings, and that a lightweight, voxel-based pocket predictor can supply that precision to a standard rigid-receptor docking engine. RAPID-Net is a five-model ensemble of 3D U-Net-like convolutional networks trained on sc-PDB cavity labels; each model outputs a soft occupancy mask for 2 Å voxels, with a ReLU in the final layer and a soft Dice loss so the network learns pocket interiors versus boundaries rather than a binary yes/no. The predicted pockets are converted into search grids centered on the pocket, with size thresholds that expand to accommodate large ligands, and AutoDock Vina performs targeted docking on each grid. On PoseBusters, this scheme yields 54.9% Top-1 PB-valid poses versus 49.1% for DiffBindFR; on the PoseBench time split it reaches 53.1% versus 59.5% for AlphaFold 3; and in 92.2% of cases at least one sampled pose is within 2 Å RMSD, regardless of rank. The paper reads these numbers as showing that pocket localization currently outweighs receptor flexibility as the accuracy bottleneck, and that better pose reranking would be the next largest gain.

Load-bearing premise

The result rests on the assumption that pocket predictions trained and tested on ligand-bound (holo) structures will localize the right search region when the receptor is unbound or computationally predicted, where the pocket need not be pre-formed.

Editorial extensions

If this is right

  • If pocket localization is the dominant bottleneck, then upgrading the pocket predictor in any blind-docking pipeline—not only Vina-based ones—should transfer most of RAPID-Net's gain.
  • The 92.2% sampling accuracy implies an immediate ceiling: a rescoring function that selects the correct pose from the ensemble could raise Top-1 success toward that number, a gain larger than any reported pocket improvement.
  • Because RAPID-Net returns search grids rather than binding-site residues, the same predictions can be fed to flexible-receptor docking engines, which the paper argues would benefit from the same focused search.
  • Threshold-less training lets the model flag secondary and allosteric sites beyond the orthosteric pocket, so pocket-guided docking can be aimed at distal functional sites rather than only the main ligand site.
  • Large complexes that exceed co-folding model input limits, such as 8F4J, become dockable when the search space is reduced to a predicted pocket.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper does not report it, but a direct experiment is available: replacing holo receptors in PoseBusters with apo or AlphaFold-predicted structures would show how much of the 54.9% success depends on seeing the bound-state pocket.
  • If the sampling-versus-ranking gap holds, adding a learned rescoring function on top of RAPID-Net's own pockets should push Top-1 accuracy toward the 92.2% sampling ceiling, a much larger gain than further pocket refinement.
  • The minority-voted pockets, which recover allosteric and exosite regions with no direct ligand contact, could serve as a candidate generator for cryptic allosteric sites in drug discovery, a direction the paper mentions but does not develop.
  • Because grid thresholds are expanded to 15 Å for minority pockets, part of the docking success may come from generous search boxes; ablating threshold sizes per pocket would isolate how much of the gain comes from the precise pocket versus the enlarged grid.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper introduces RAPID-Net, a five-member ensemble of 3D U-Net-style convolutional networks for protein pocket prediction, trained on cavity labels from sc-PDB with soft Dice loss and ReLU output. Predictions are converted into AutoDock Vina search grids using hand-set expansion thresholds, and docking is evaluated on PoseBusters, Astex, Coach420, and BU48. The central claim is that precise pocket identification is the decisive factor in blind docking accuracy: RAPID-Net-guided Vina achieves 54.9% Top-1 PoseBusters-valid poses on PoseBusters versus 49.1% for DiffBindFR, 53.1% versus 59.5% for AlphaFold 3 on the PoseBench time split, and 92.2% sampling accuracy (at least one pose with RMSD<2 Å). The paper also reports per-run variability, PLI metrics, and qualitative allosteric-site case studies.

Significance. If the empirical results withstand scrutiny, the paper makes a useful practical contribution: a lightweight pocket predictor that can be plugged into standard docking pipelines, with public code and reproducible notebooks. The explicit separation of Top-1 accuracy from sampling accuracy is a valuable diagnostic, and the comparison against recent blind-docking and co-folding tools is informative. The main significance depends on the integrity of the benchmark evaluation, because the headline comparisons are the primary evidence for the claim that pocket identification, rather than pose ranking or receptor flexibility, is the bottleneck.

major comments (4)
  1. [Sections VII–VIII, Tables I–II] The manuscript reports no train/test overlap analysis for the PoseBusters and Astex benchmarks, although sc-PDB (Ref. 81) is the training set and Astex (Ref. 33) is a curated PDB-derived set. Section IX explicitly excludes training-set structures for Coach420 and BU48, but no equivalent exclusion or check is described for PoseBusters or Astex. If any Astex or older PoseBusters complexes appear in sc-PDB, the reported Top-1 accuracies, RMSD distributions, and PLI values in Sections VII–VIII would be inflated by memorization. The PoseBench time split (deposited after 30 September 2021) is less exposed if the training snapshot predates it, but the headline 54.9% PoseBusters number and the Astex numbers need a concrete overlap analysis or a revised, more cautious claim.
  2. [Section V, Fig. 5] The docking protocol depends on hand-set grid expansion thresholds of 2/5 Å for majority-voted pockets and 2/5/10/15 Å for minority-reported pockets. The 8FAV example in Fig. 5 shows that docking can succeed even when no predicted pocket overlaps the true ligand pose, purely because the expanded search grid covers the site. The paper does not report an ablation varying these thresholds or a control with a fixed large grid, so the central claim that precise pocket identification is the decisive driver of docking success is not cleanly separated from the effect of grid enlargement. Please report results for at least one alternative threshold set and clearly state the sensitivity of the headline metrics to this protocol choice.
  3. [Sections V and VII, Fig. 6] The comparison to DiffBindFR and AlphaFold 3 confounds pocket quality with the docking engine and protocol. No unguided AutoDock Vina baseline (docking over the whole protein with the same Exhaustiveness and num_modes settings) is reported. Since the central claim is that pocket identification drives docking success, the reader needs to see RAPID-Net-guided Vina versus unguided Vina under otherwise identical settings; without this baseline, part of the 54.9% result could reflect Vina's search behavior rather than RAPID-Net's pockets. The 'prior knowledge' Vina result of 93.8% sampling accuracy in Section VII is not an adequate substitute because it uses the true ligand coordinates to define the search box.
  4. [Section VII, Figs. 6 and 8] The headline differences (54.9% vs 49.1% and 53.1% vs 59.5%) are reported without confidence intervals, bootstrap estimates, or repeated-run statistics at the docking level. The five RAPID-Net runs in Tables I–IV and Figs. 6 and 18 show substantial run-to-run variability in pocket coverage and PLI, so the ensemble docking accuracy may also vary with the random seeds or training runs. Please provide variability estimates for the primary Top-1 and sampling-accuracy metrics, or at least state explicitly whether the docking results were obtained from a single ensemble checkpoint.
minor comments (6)
  1. [Eqs. (4)–(5)] The notation max(x,y,z) and min(x,y,z) is ambiguous: these should be coordinate-wise extrema over the pocket atoms, not a scalar maximum or minimum of a single triple. Please clarify by writing e.g. x_max = max_i x_i, with analogous expressions for y and z.
  2. [Section VI and Fig. 7] Section VI says the Top-1 RMSD is computed 'between the predicted and one of the true ligand poses if multiple true poses are available,' while Fig. 7's caption says RMSD to the closest one is reported. Please make the metric definition consistent throughout.
  3. [Section X and Fig. 27] The descriptive claims about allosteric sites, exosites, and bridges are qualitative and based on visual inspection; please add a quantitative measure, such as residue-level overlap with annotated exosite or allosteric-site residues, to support these statements.
  4. [Section VI] The paper states in the introduction that all evaluations are performed on holo structures, but this important limitation is not restated in the metrics or benchmark sections; please state it explicitly alongside the PoseBusters and Astex results.
  5. [References] Reference [85] contains the placeholder 'Accessed: YYYY-MM-DD'; please replace it with the actual access date.
  6. [Fig. 27 caption] The term 'minimally-reported pockets' in the Fig. 27 caption should be 'minority-reported pockets' for consistency with the rest of the text.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the headline results are external docking benchmarks, not quantities derived from fitted parameters or self-cited theorems.

full rationale

The paper's derivation chain is not circular. RAPID-Net is trained on sc-PDB cavity labels and then evaluated on PoseBusters, Astex, Coach420, and BU48 through docking with AutoDock Vina and PLI metrics. None of the reported success rates are tautological: the docking accuracy numbers depend on Vina's scoring and sampling as well as on the predicted pockets, and the grid thresholds (2, 5, 10, 15 Å) are fixed protocol choices rather than parameters fitted to the benchmarks. The central claim that pocket localization is a decisive driver of docking success is an empirical conclusion drawn from external comparisons, not a definitional equivalence. The only self-citation is reference [80] supporting the use of a single attention block; the paper also states 'Our experiments similarly suggest' the same conclusion, so the self-citation is not load-bearing. The disclosed limitation that evaluations are performed on holo structures, and the lack of an explicit train/test overlap analysis for PoseBusters and Astex, are correctness and generalization risks rather than circularity. No fitted value is renamed as a prediction, and no uniqueness or ansatz is imported via self-citation.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

No invented physical entities are introduced. The model is trained on sc-PDB cavity labels, and the protocol depends on hand-chosen thresholds and grid expansions. The most consequential assumptions are that holo structures represent blind docking targets and that Vina's default settings are an adequate docking engine.

free parameters (4)
  • Inference threshold = 0.5
    A voxel is called part of a pocket if the network output exceeds 0.5; chosen by hand with no sensitivity analysis reported in Section IV.
  • Ensemble voting threshold = 3 of 5 models
    Majority-voted pockets require at least 3 of 5 replicas; this balances precision and recall but is not optimized on a held-out set in Section III.
  • Search-grid expansion thresholds = 2, 5, 10, 15 Angstroms
    Majority pockets use 2 and 5 Angstroms; minority pockets use 2, 5, 10, and 15 Angstroms. These hand-set values directly control the Vina search box and docking success in Section V.
  • Ensemble size = 5
    Five independently trained replicas are aggregated; the number is chosen by hand and affects the stability and coverage metrics in Section III.
assumptions (4)
  • domain assumption sc-PDB VolSite cavity pseudoatoms, using threshold-less cavityALL.mol2 labels, define the complete set of relevant binding pockets.
    The model is trained and evaluated on these labels; the assumption that distal and allosteric sites are captured by threshold-less labels is central to the paper's claims in Section IV.
  • domain assumption Holo protein structures with bound ligands are a valid testbed for binding-site-agnostic docking, and performance transfers to apo or predicted structures.
    All docking evaluations are performed on holo structures with rigid-receptor Vina, as stated in Section V and the introduction; real screening often starts from apo or predicted structures.
  • domain assumption AutoDock Vina with default settings (exhaustiveness 32, num_modes 40) is a sufficient docking engine, so the search grid is the only guided variable.
    The protocol in Section V fixes Vina defaults, so the reported docking accuracies are attributed to pocket guidance rather than to Vina tuning.
  • domain assumption The 18 tfbio voxel features and the soft Dice loss provide a sufficient representation and training signal for pocket geometry.
    Section IV specifies tfbio features and the soft segmentation loss; if these omit relevant chemistry, the model's generalization would be limited.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RAPID-Net: Accurate Pocket Identification for Binding-Site-Agnostic Docking." pith.science (2026). https://pith.science/paper/NZSDBSOQ

@misc{pith2026250202371,
  author       = {Pith},
  title        = {Pith review of: RAPID-Net: Accurate Pocket Identification for Binding-Site-Agnostic Docking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NZSDBSOQ}},
  note         = {Machine review of arXiv:2502.02371}
}
read the original abstract

Accurate identification of druggable pockets and their features is essential for structure-based drug design and effective downstream docking. Here, we present RAPID-Net, a deep learning-based algorithm designed for the accurate prediction of binding pockets and seamless integration with docking pipelines. On the PoseBusters benchmark, RAPID-Net-guided AutoDock Vina achieves 54.9% of Top-1 poses with RMSD < 2 A and satisfying the PoseBusters chemical-validity criterion, compared to 49.1% for DiffBindFR. On the most challenging time split of PoseBusters aiming to assess generalization ability (structures submitted after September 30, 2021), RAPID-Net-guided AutoDock Vina achieves 53.1% of Top-1 poses with RMSD < 2 A and PB-valid, versus 59.5% for AlphaFold 3. Notably, in 92.2% of cases, RAPID-Net-guided Vina samples at least one pose with RMSD < 2 A (regardless of its rank), indicating that pose ranking, rather than sampling, is the primary accuracy bottleneck. The lightweight inference, scalability, and competitive accuracy of RAPID-Net position it as a viable option for large-scale virtual screening campaigns. Across diverse benchmark datasets, RAPID-Net outperforms other pocket prediction tools, including PUResNet and Kalasanty, in both docking accuracy and pocket-ligand intersection rates. Furthermore, we demonstrate the potential of RAPID-Net to accelerate the development of novel therapeutics by highlighting its performance on pharmacologically relevant targets. RAPID-Net accurately identifies distal functional sites, offering new opportunities for allosteric inhibitor design. In the case of the RNA-dependent RNA polymerase of SARS-CoV-2, RAPID-Net uncovers a wider array of potential binding pockets than existing predictors, which typically annotate only the orthosteric pocket and overlook secondary cavities.

Figures

Figures reproduced from arXiv: 2502.02371 by the authors.

Figure 1
Figure 1. 8DP2 protein structure from the PoseBusters [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. For the 8F4J protein structure from the PoseBusters [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Schematic representation of RAPID-Net (ReLU-Activated Pocket Identification for Docking). Key improvements that distinguish RAPID-Net from previous approaches include a ReLU activation in the final layer, the usage of a soft dice loss function, including a single SE-attention block, and removing redundant residual connections [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (22 more)
Figure 4
Figure 4. Figure 4: Majority-voted pocket predicted by RAPID-Net and [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: For the 8FAV protein structure, our model predictions [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Comparison of Top-1 Vina accuracies when guided by [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 8
Figure 8. Figure 8: Performance comparison on the subset of novel [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 7
Figure 7. Figure 7: Distribution of RMSD values for Top-1 Vina predicted [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 9
Figure 9. Figure 9: In 7KZ9 protein structure, RAPID-Net predicts [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 12
Figure 12. Figure 12: Solid bars represent Top-1 accuracies. Dashed seg [PITH_FULL_IMAGE:figures/full_fig_p009_12.png]
Figure 11
Figure 11. Figure 11: For the 7XFA protein structure, neither the Top-1 nor [PITH_FULL_IMAGE:figures/full_fig_p009_11.png]
Figure 13
Figure 13. Figure 13: For the 7PRI protein structure, the Top-1 Vina pose fails while a subleading one succeeds. In [PITH_FULL_IMAGE:figures/full_fig_p010_13.png]
Figure 15
Figure 15. Figure 15: For the 7NUT protein structure, there are two true [PITH_FULL_IMAGE:figures/full_fig_p010_15.png]
Figure 14
Figure 14. Figure 14: In the 7P1F protein structure, two true ligand binding [PITH_FULL_IMAGE:figures/full_fig_p010_14.png]
Figure 16
Figure 16. Figure 16: In the 7A9E protein structure, RAPID-Net predicts a [PITH_FULL_IMAGE:figures/full_fig_p011_16.png]
Figure 17
Figure 17. Figure 17: Distribution of the maximum PLI scores correspond [PITH_FULL_IMAGE:figures/full_fig_p011_17.png]
Figure 20
Figure 20. Figure 20: Distribution of the maximum PLI scores correspond [PITH_FULL_IMAGE:figures/full_fig_p012_20.png]
Figure 21
Figure 21. Figure 21: Docking accuracy of AutoDock Vina 12 for the Astex Diverse Set 33 when guided by different pocket prediction algorithms, comparing Top-1 and ensemble accuracy. Method Nonzero Within 15 A PLI ˚ Kalasanty 73 70 78.72% PUResNet 76 72 82.36% RAPID Run 1 85 83 94.83% RAPID…
Figure 22
Figure 22. Figure 22: In the 1G9V protein structure from Astex Diverse [PITH_FULL_IMAGE:figures/full_fig_p013_22.png]
Figure 23
Figure 23. Figure 23: Distribution of the maximum PLI scores correspond [PITH_FULL_IMAGE:figures/full_fig_p013_23.png]
Figure 24
Figure 24. Figure 24: Distribution of the maximum PLI scores correspond [PITH_FULL_IMAGE:figures/full_fig_p013_24.png]
Figure 25
Figure 25. Figure 25: For the Nsp12 protein64, the pockets predicted by Kalasanty70 and PUResNet 69 – shown as green and orange dots, respectively – and the likely interacting residues predicted by PUResNet V266 correspond to the orthosteric remdesivir binding site 64. In contrast, RAPID-N…
Figure 26
Figure 26. Figure 26: Thrombin (RCSB PDB: 1DWC). Unlike the pockets predicted by Kalasanty70 and PUResNet 69 –shown as green and orange dots, respectively – the majority-voted pocket predicted by our model, shown as cyan sticks, has a distinct bulge extending toward residues associated wit…
Figure 27
Figure 27. Figure 27: Majority-voted pockets predicted by RAPID-Net are represented by cyan sticks, while the minimally-reported pockets [PITH_FULL_IMAGE:figures/full_fig_p016_27.png]
Figure 28
Figure 28. Figure 28: Ls-AChBP (RCSB PDB: 2ZJV). The majority-voted pocket predicted by our model has a distinct bridge towards the Thr155-Asp160 region belonging to the F loop. Kalasanty70 and PUResNet V169 do not predict any pockets for this protein structure, while PUResNet V266 predict…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

126 extracted references · 78 canonical work pages

  1. [1]

    Kukol, A.; othersMolecular modeling of proteins; Springer, 2008; V ol. 443

    2008

  2. [2]

    druggable

    An, J.; Totrov, M.; Abagyan, R. Comprehensive iden- tification of “druggable” protein ligand binding sites. Genome Informatics2004,15, 31–41. 3https://github.com/BalytskyiJaroslaw/RAPID-Net

  3. [3]

    De Ruyck, J.; Brysbaert, G.; Blossey, R.; Lensink, M. F. Molecular docking as a popular tool in drug design, an in silico travel.Advances and Applications in Bioinfor- matics and Chemistry2016, 1–11

  4. [4]

    A.; Li, Y .; Wu, D.; Li, X.; Li, C

    Rezaei, M. A.; Li, Y .; Wu, D.; Li, X.; Li, C. Deep learning in drug design: protein-ligand binding affinity prediction.IEEE/ACM transactions on computational biology and bioinformatics2020,19, 407–417

  5. [5]

    Vamathevan, J.; Clark, D.; Czodrowski, P.; Dunham, I.; Ferran, E.; Lee, G.; Li, B.; Madabhushi, A.; Shah, P.; Spitzer, M.; others Applications of machine learning in drug discovery and development.Nature reviews Drug discovery2019,18, 463–477

  6. [6]

    Molecular docking: challenges, advances and its use in drug discovery perspective

    Saikia, S.; Bordoloi, M. Molecular docking: challenges, advances and its use in drug discovery perspective. Current drug targets2019,20, 501–521

  7. [7]

    Docking, virtual high throughput screening and in silico fragment-based drug design.Journal of cellular and molecular medicine 2009,13, 238–248

    Zoete, V .; Grosdidier, A.; Michielin, O. Docking, virtual high throughput screening and in silico fragment-based drug design.Journal of cellular and molecular medicine 2009,13, 238–248

    2009

  8. [8]

    Structure-based virtual screening for drug discovery: principles, applications and recent advances.Current topics in medicinal chemistry2014,14, 1923–1938

    Lionta, E.; Spyrou, G.; K Vassilatis, D.; Cournia, Z. Structure-based virtual screening for drug discovery: principles, applications and recent advances.Current topics in medicinal chemistry2014,14, 1923–1938

    1923

Show all 126 references
  1. [9]

    L.An introduction to medicinal chemistry; Oxford university press, 2023

    Patrick, G. L.An introduction to medicinal chemistry; Oxford university press, 2023

    2023

  2. [10]

    Y .; Al- tuzar, V .; Vivanco-Cid, H.; Mendoza-Barrera, C

    Hern ´andez-Santoyo, A.; Tenorio-Barajas, A. Y .; Al- tuzar, V .; Vivanco-Cid, H.; Mendoza-Barrera, C. Protein-protein and protein-ligand docking.Protein engineering-technology and application2013, 63–81

  3. [11]

    Trott, O.; Olson, A. J. AutoDock Vina: improving the speed and accuracy of docking with a new scoring func- tion, efficient optimization, and multithreading.Journal of computational chemistry2010,31, 455–461

  4. [12]

    F.; Forli, S

    Eberhardt, J.; Santos-Martins, D.; Tillack, A. F.; Forli, S. AutoDock Vina 1.2. 0: New docking methods, expanded force field, and python bindings.Journal of chemical information and modeling2021,61, 3891– 3898

  5. [13]

    L.; Cole, J

    Verdonk, M. L.; Cole, J. C.; Hartshorn, M. J.; Mur- ray, C. W.; Taylor, R. D. Improved protein–ligand docking using GOLD.Proteins: Structure, Function, and Bioinformatics2003,52, 609–623

  6. [14]

    A.; Banks, J

    Friesner, R. A.; Banks, J. L.; Murphy, R. B.; Hal- gren, T. A.; Klicic, J. J.; Mainz, D. T.; Repasky, M. P.; Knoll, E. H.; Shelley, M.; Perry, J. K.; others Glide: a new approach for rapid, accurate docking and scoring

  7. [15]

    Method and assessment of docking accuracy.Journal of medicinal chemistry2004,47, 1739–1749

  8. [16]

    Grasso, G.; Di Gregorio, A.; Mavkov, B.; Piga, D.; Labate, G. F. D.; Danani, A.; Deriu, M. A. Fragmented blind docking: a novel protein–ligand binding predic- tion protocol.Journal of Biomolecular Structure and Dynamics2022,40, 13472–13481

  9. [17]

    M.; Alhossary, A

    Hassan, N. M.; Alhossary, A. A.; Mu, Y .; Kwoh, C.- K. Protein-ligand blind docking using QuickVina-W with inter-process spatio-temporal integration.Scientific reports2017,7, 15451

  10. [18]

    R.; Baumgartner, M

    Koes, D. R.; Baumgartner, M. P.; Camacho, C. J. Lessons learned in empirical scoring with smina from the CSAR 2011 benchmarking exercise.Journal of chemical information and modeling2013,53, 1893– 1904

    2011

  11. [19]

    Predicting protein structure from single se- quences.Nature Computational Science2022,2, 775– 776

    Shen, Y . Predicting protein structure from single se- quences.Nature Computational Science2022,2, 775– 776

  12. [20]

    nature2021,596, 583–589

    Jumper, J.; Evans, R.; Pritzel, A.; Green, T.; Fig- urnov, M.; Ronneberger, O.; Tunyasuvunakool, K.; Bates, R.; ˇZ´ıdek, A.; Potapenko, A.; others Highly accurate protein structure prediction with AlphaFold. nature2021,596, 583–589

  13. [21]

    ColabFold: making protein folding accessible to all.Nature methods2022, 19, 679–682

    Mirdita, M.; Sch ¨utze, K.; Moriwaki, Y .; Heo, L.; Ovchinnikov, S.; Steinegger, M. ColabFold: making protein folding accessible to all.Nature methods2022, 19, 679–682

  14. [22]

    Ahdritz, G.; Bouatta, N.; Floristean, C.; Kadyan, S.; Xia, Q.; Gerecke, W.; O’Donnell, T. J.; Berenberg, D.; Fisk, I.; Zanichelli, N.; others OpenFold: Retraining AlphaFold2 yields new insights into its learning mech- anisms and capacity for generalization.Nature Methods 2024, 1–11

    2024

  15. [23]

    R.; Wang, J.; Cong, Q.; Kinch, L

    Baek, M.; DiMaio, F.; Anishchenko, I.; Dauparas, J.; Ovchinnikov, S.; Lee, G. R.; Wang, J.; Cong, Q.; Kinch, L. N.; Schaeffer, R. D.; others Accurate predic- tion of protein structures and interactions using a three- track neural network.Science2021,373, 871–876

  16. [24]

    P.; Al-Lazikani, B.; Hopkins, A

    Overington, J. P.; Al-Lazikani, B.; Hopkins, A. L. How many drug targets are there?Nature reviews Drug discovery2006,5, 993–996

  17. [25]

    P.; Donadi, R

    Santos, R.; Ursu, O.; Gaulton, A.; Bento, A. P.; Donadi, R. S.; Bologa, C. G.; Karlsson, A.; Al- Lazikani, B.; Hersey, A.; Oprea, T. I.; others A compre- hensive map of molecular drug targets.Nature reviews Drug discovery2017,16, 19–34

  18. [26]

    Huang, Y .; Zhang, H.; Jiang, S.; Yue, D.; Lin, X.; Zhang, J.; Gao, Y . Q. DSDP: a blind docking strategy accelerated by GPUs.Journal of Chemical Information and Modeling2023,63, 4355–4363

  19. [27]

    S.; Barton, G

    Utg ´es, J. S.; Barton, G. J. Comparative evaluation of methods for the prediction of protein–ligand binding sites.Journal of Cheminformatics2024,16, 126

  20. [28]

    DiffBindFR: an SE(3) equivariant network for flexible protein-ligand docking

    Zhu, J.; Gu, Z.; Pei, J.; Lai, L. DiffBindFR: an SE(3) equivariant network for flexible protein-ligand docking. Chemical Science2024,15, 7926–7942, Supplementary information available: PDF (1982K)

  21. [29]

    M.; Deane, C

    Buttenschoen, M.; Morris, G. M.; Deane, C. M. Pose- Busters: AI-based docking methods fail to generate physically valid poses or generalise to novel sequences. Chemical Science2024,15, 3130–3139

  22. [30]

    Dong, C.; Huang, Y .-P.; Lin, X.; Zhang, H.; Gao, Y . Q. DSDPFlex: Flexible-Receptor Docking with GPU Ac- celeration.Journal of Chemical Information and Mod- eling2024,64, 5252–5265

  23. [31]

    Abramson, J.; Adler, J.; Dunger, J.; Evans, R.; Green, T.; Pritzel, A.; Jumper, J. M. Accurate structure prediction of biomolecular interactions with AlphaFold 3.Nature2024, 1–3

  24. [32]

    Deep Learning for Protein-Ligand Docking: Are We There Yet? ICML AI4Science Workshop

    Morehead, A.; Giri, N.; Liu, J.; Neupane, P.; Cheng, J. Deep Learning for Protein-Ligand Docking: Are We There Yet? ICML AI4Science Workshop. 2024; selected as a spotlight presentation

    2024

  25. [33]

    Chollet, F.Deep learning with Python; Simon and Schuster, 2021

    2021

  26. [34]

    J.; Verdonk, M

    Hartshorn, M. J.; Verdonk, M. L.; Chessari, G.; Brew- erton, S. C.; Mooij, W. T.; Mortenson, P. N.; Mur- ray, C. W. Diverse, high-quality test set for the vali- dation of protein- ligand docking performance.Journal of Medicinal Chemistry2007,50, 726–741

  27. [35]

    COFACTOR: An Accu- rate Comparative Algorithm for Structure-Based Protein Function Annotation.Nucleic Acids Res.2012,40, W471–W477

    Roy, A.; Yang, J.; Zhang, Y . COFACTOR: An Accu- rate Comparative Algorithm for Structure-Based Protein Function Annotation.Nucleic Acids Res.2012,40, W471–W477

    2012

  28. [36]

    LIGSITE csc: predicting ligand binding sites using the Connolly surface and degree of conservation.BMC Structural Biology2006, 6, 1–11

    Huang, B.; Schroeder, M. LIGSITE csc: predicting ligand binding sites using the Connolly surface and degree of conservation.BMC Structural Biology2006, 6, 1–11

  29. [37]

    Fpocket: An Open Source Platform for Ligand Pocket Detection

    Le Guilloux, V .; Schmidtke, P.; Tuff ´ery, P. Fpocket: An Open Source Platform for Ligand Pocket Detection. BMC Bioinformatics2009,10, 168

  30. [38]

    LIGSITE: Automatic and Efficient Detection of Potential Small Molecule-Binding Sites in Proteins.J

    Hendlich, M.; Rippmann, F.; Barnickel, G. LIGSITE: Automatic and Efficient Detection of Potential Small Molecule-Binding Sites in Proteins.J. Mol. Graph. Model.1997,15, 359–363, 389

    1997

  31. [39]

    Laskowski, R. A. SURFNET: A Program for Visual- izing Molecular Surfaces, Cavities, and Intermolecular Interactions.J. Mol. Graph.1995,13, 323–330, 307– 308

    1995

  32. [40]

    G.; Banaszak, L

    Levitt, D. G.; Banaszak, L. J. POCKET: A Computer Graphics Method for Identifying and Displaying Protein Cavities and Their Surrounding Amino Acids.J. Mol. Graph.1992,10, 229–234

    1992

  33. [41]

    J.; Jones, T

    Kleywegt, G. J.; Jones, T. A. Detection, Delineation, Measurement and Display of Cavities in Macromolec- ular Structures.Acta Crystallogr., Sect. D: Biol. Crys- tallogr.1994,50, 178–185

    1994

  34. [42]

    Anatomy of Protein Pockets and Cavities: Measurement of Binding Site Geometry and Implications for Ligand Design

    Liang, J.; Edelsbrunner, H.; Woodward, C. Anatomy of Protein Pockets and Cavities: Measurement of Binding Site Geometry and Implications for Ligand Design. Protein Sci.1998,7, 1884–1897

    1998

  35. [43]

    Brady, G. P. J.; Stouten, P. F. Fast Prediction and Visualization of Protein Binding Pockets with PASS. J. Comput.-Aided Mol. Des.2000,14, 383–401

    2000

  36. [44]

    PocketPicker: Analysis of Ligand Binding-Sites with Shape Descrip- tors.Chem

    Weisel, M.; Proschak, E.; Schneider, G. PocketPicker: Analysis of Ligand Binding-Sites with Shape Descrip- tors.Chem. Cent. J.2007,1, 7

    2007

  37. [45]

    Pocketome via Com- prehensive Identification and Classification of Ligand Binding Envelopes.Mol

    An, J.; Totrov, M.; Abagyan, R. Pocketome via Com- prehensive Identification and Classification of Ligand Binding Envelopes.Mol. Cell. Proteomics2005,4, 752– 761

  38. [46]

    Goodford, P. J. A Computational Procedure for Deter- mining Energetically Favorable Binding Sites on Bi- ologically Important Macromolecules.J. Med. Chem. 1985,28, 849–857

    1985

  39. [47]

    Druggable

    An, J.; Totrov, M.; Abagyan, R. Comprehensive Iden- tification of “Druggable” Protein Ligand Binding Sites. Genome Inform.2004,15, 31–41

    2004

  40. [48]

    Laurie, A. T. R.; Jackson, R. M. Q-SiteFinder: An Energy-Based Method for the Prediction of Pro- tein–Ligand Binding Sites.Bioinformatics2005,21, 1908–1916

    1908

  41. [49]

    EasyMIFS and SiteHound: A Toolkit for the Identification of Ligand-Binding Sites in Protein Structures.Bioinformatics2009,25, 3185– 3186

    Ghersi, D.; Sanchez, R. EasyMIFS and SiteHound: A Toolkit for the Identification of Ligand-Binding Sites in Protein Structures.Bioinformatics2009,25, 3185– 3186

  42. [50]

    E.; Beglov, D.; Villar, E

    Ngan, C.-H.; Bohnuud, T.; Mottarella, S. E.; Beglov, D.; Villar, E. A.; Hall, D. R.; Vajda, S. FTSite: High Ac- curacy Detection of Ligand Binding Sites on Unbound Protein Structures.Bioinformatics2012,28, 286–287

  43. [51]

    ConSurf: An Algo- rithmic Tool for the Identification of Functional Regions in Proteins by Surface Mapping of Phylogenetic Infor- mation.J

    Armon, A.; Graur, D.; Ben-Tal, N. ConSurf: An Algo- rithmic Tool for the Identification of Functional Regions in Proteins by Surface Mapping of Phylogenetic Infor- mation.J. Mol. Biol.2001,307, 447–463

    2001

  44. [52]

    E.; Mayrose, I.; Glaser, F.; Ben- Tal, N

    Pupko, T.; Bell, R. E.; Mayrose, I.; Glaser, F.; Ben- Tal, N. Rate4Site: An Algorithmic Tool for the Identi- fication of Functional Regions in Proteins by Surface Mapping of Evolutionary Determinants Within Their Homologues.Bioinformatics2002,18, S71–S77

  45. [53]

    Ligand-Binding Site Pre- diction Using Ligand-Interacting and Binding Site- Enriched Protein Triangles.Bioinformatics2012,28, 1579–1585

    Xie, Z.-R.; Hwang, M.-J. Ligand-Binding Site Pre- diction Using Ligand-Interacting and Binding Site- Enriched Protein Triangles.Bioinformatics2012,28, 1579–1585

  46. [54]

    J.; Barton, G

    Zvelebil, M. J.; Barton, G. J.; Taylor, W. R.; Stern- berg, M. J. E. Prediction of Protein Secondary Structure and Active Sites Using the Alignment of Homologous Sequences.J. Mol. Biol.1987,195, 957–961

    1987

  47. [55]

    N.; Kelley, L

    Wass, M. N.; Kelley, L. A.; Sternberg, M. J. E. 3DLi- gandSite: Predicting Ligand-Binding Sites Using Sim- ilar Structures.Nucleic Acids Res.2010,38, W469– W473

    2010

  48. [56]

    Protein–Ligand Bind- ing Site Recognition Using Complementary Binding- Specific Substructure Comparison and Sequence Profile Alignment.Bioinformatics2013,29, 2588–2595

    Yang, J.; Roy, A.; Zhang, Y . Protein–Ligand Bind- ing Site Recognition Using Complementary Binding- Specific Substructure Comparison and Sequence Profile Alignment.Bioinformatics2013,29, 2588–2595

  49. [57]

    Ligand Binding Site Detection by Local Structure Alignment and Its Performance Com- plementarity.J

    Lee, H.-S.; Im, W. Ligand Binding Site Detection by Local Structure Alignment and Its Performance Com- plementarity.J. Chem. Inf. Model.2013,53, 2462–2470

    2013

  50. [58]

    Brylinski, M.; Feinstein, W. P. eFindSite: Improved Prediction of Ligand Binding Sites in Protein Models Using Meta-Threading, Machine Learning and Auxil- iary Ligands.J. Comput.-Aided Mol. Des.2013,27, 551–567

    2013

  51. [59]

    E.; Bechor- Shental, D.; Martz, E.; Ben-Tal, N

    Glaser, F.; Pupko, T.; Paz, I.; Bell, R. E.; Bechor- Shental, D.; Martz, E.; Ben-Tal, N. A Method for Lo- calizing Ligand Binding Pockets in Protein Structures. Proteins2006,62, 479–488

  52. [60]

    Halgren, T. A. Identifying and Characterizing Binding Sites and Assessing Druggability.J. Chem. Inf. Model. 2009,49, 377–389

    2009

  53. [61]

    A.; Laskowski, R

    Capra, J. A.; Laskowski, R. A.; Thornton, J. M.; Singh, M.; Funkhouser, T. A. Predicting Protein Ligand Binding Sites by Combining Evolutionary Sequence Conservation and 3D Structure.PLoS Comput. Biol. 2009,5, e1000585

    2009

  54. [62]

    MetaPocket: A Meta Approach to Improve Protein Ligand Binding Site Prediction.OMICS2009, 13, 325–330

    Huang, B. MetaPocket: A Meta Approach to Improve Protein Ligand Binding Site Prediction.OMICS2009, 13, 325–330

  55. [63]

    D.; Blundell, T

    Bray, T.; Marsden, B. D.; Blundell, T. L. SitesIdentify: A Protein Functional Site Prediction Tool.BMC Bioin- formatics2009,10, 379

  56. [64]

    FINDSITE: A Threading- Based Approach to Ligand Homology Modeling.PLoS Comput

    Brylinski, M.; Skolnick, J. FINDSITE: A Threading- Based Approach to Ligand Homology Modeling.PLoS Comput. Biol.2009,5, e1000405

    2009

  57. [65]

    Science2020,368, 1499–1504

    Yin, W.; Mao, C.; Luan, X.; Shen, D.-D.; Shen, Q.; Su, H.; Wang, X.; Zhou, F.; Zhao, W.; Gao, M.; others Structural basis for inhibition of the RNA-dependent RNA polymerase from SARS-CoV-2 by remdesivir. Science2020,368, 1499–1504

  58. [66]

    R.; Studer, G.; Tauriello, G.; Water- house, A

    Gervasoni, S.; Vistoli, G.; Talarico, C.; Manelfi, C.; Beccari, A. R.; Studer, G.; Tauriello, G.; Water- house, A. M.; Schwede, T.; Pedretti, A. A compre- hensive mapping of the druggable cavities within the SARS-CoV-2 therapeutically relevant proteins by com- bining pocket an...

    2020

  59. [67]

    Jeevan, K.; Palistha, S.; Tayara, H.; Chong, K. T. PUResNetV2.0: a deep learning model leveraging sparse representation for improved ligand binding site prediction.Journal of Cheminformatics2024,16, 1–16

  60. [68]

    VN-EGNN: E (3)- Equivariant Graph Neural Networks with Virtual Nodes Enhance Protein Binding Site Identification.arXiv preprint arXiv:2404.071942024,

    Sestak, F.; Schneckenreiter, L.; Brandstetter, J.; Hochre- iter, S.; Mayr, A.; Klambauer, G. VN-EGNN: E (3)- Equivariant Graph Neural Networks with Virtual Nodes Enhance Protein Binding Site Identification.arXiv preprint arXiv:2404.071942024,

    arXiv

  61. [69]

    P.; Tiwary, P

    Smith, Z.; Strobel, M.; Vani, B. P.; Tiwary, P. Graph attention site prediction (grasp): Identifying druggable binding sites using graph neural networks with atten- tion.Journal of chemical information and modeling 2024,64, 2637–2644

    2024

  62. [70]

    Kandel, J.; Tayara, H.; Chong, K. T. PUResNet: predic- tion of protein-ligand binding sites using deep residual neural network.Journal of cheminformatics2021,13, 1–14

  63. [71]

    M.; Zielenkiewicz, P.; Siedlecki, P

    Stepniewska-Dziubinska, M. M.; Zielenkiewicz, P.; Siedlecki, P. Improving detection of protein-ligand bind- ing sites with 3D segmentation.Scientific Reports2020, 10, 5035

  64. [72]

    Improving protein-ligand bind- ing site prediction accuracy by classification of inner pocket points using local features.Journal of chemin- formatics2015,7, 1–13

    Kriv ´ak, R.; Hoksza, D. Improving protein-ligand bind- ing site prediction accuracy by classification of inner pocket points using local features.Journal of chemin- formatics2015,7, 1–13

  65. [73]

    Aggarwal, R.; Gupta, A.; Chelur, V .; Jawahar, C.; Priyakumar, U. D. DeepPocket: ligand binding site detection and segmentation using 3D convolutional neural networks.Journal of Chemical Information and Modeling2021,62, 5069–5079

  66. [74]

    Fpocket: an open source platform for ligand pocket detection.BMC bioinformatics2009,10, 1–11

    Le Guilloux, V .; Schmidtke, P.; Tuffery, P. Fpocket: an open source platform for ligand pocket detection.BMC bioinformatics2009,10, 1–11

  67. [75]

    Carbery, A.; Buttenschoen, M.; Skyner, R.; von Delft, F.; Deane, C. M. Learnt representations of pro- teins can be used for accurate prediction of small molecule binding sites on experimentally determined and predicted protein structures.Journal of Cheminfor- matics2024,16, 32

  68. [76]

    PGBind: pocket-guided explicit attention learning for protein– ligand docking.Briefings in Bioinformatics2024,25, bbae455

    Shen, A.; Yuan, M.; Ma, Y .; Du, J.; Wang, M. PGBind: pocket-guided explicit attention learning for protein– ligand docking.Briefings in Bioinformatics2024,25, bbae455

  69. [77]

    U-Net: Convo- lutional networks for biomedical image segmentation

    Ronneberger, O.; Fischer, P.; Brox, T. U-Net: Convo- lutional networks for biomedical image segmentation. arXiv preprint arXiv:1505.045972015,

    arXiv

  70. [78]

    Deep residual learning for image recognition

    He, K.; Zhang, X.; Ren, S.; Sun, J. Deep residual learning for image recognition. Proceedings of the IEEE conference on computer vision and pattern recognition. 2016; pp 770–778

    2016

  71. [79]

    Squeeze-and-excitation net- works

    Hu, J.; Shen, L.; Sun, G. Squeeze-and-excitation net- works. Proceedings of the IEEE conference on com- puter vision and pattern recognition. 2018; pp 7132– 7141

    2018

  72. [80]

    Breast cancer histopathological image classification using convolu- tional neural networks with small SE-ResNet module

    Jiang, Y .; Chen, L.; Zhang, H.; Xiao, X. Breast cancer histopathological image classification using convolu- tional neural networks with small SE-ResNet module. PloS One2019,14, e0214587

  73. [81]

    Enhancing Open-World Bacterial Raman Spectra Identification by Feature Regularization for Im- proved Resilience against Unknown Classes.Chemical & Biomedical Imaging2024,2, 442–452

    Balytskyi, Y .; Kalashnyk, N.; Hubenko, I.; Balytska, A.; McNear, K. Enhancing Open-World Bacterial Raman Spectra Identification by Feature Regularization for Im- proved Resilience against Unknown Classes.Chemical & Biomedical Imaging2024,2, 442–452

  74. [82]

    sc- PDB: a 3D-database of ligandable binding sites—10 years on.Nucleic Acids Research2015,43, D399– D404

    Desaphy, J.; Bret, G.; Rognan, D.; Kellenberger, E. sc- PDB: a 3D-database of ligandable binding sites—10 years on.Nucleic Acids Research2015,43, D399– D404

  75. [83]

    Comparison and druggability prediction of protein–ligand binding sites from pharmacophore- annotated cavity shapes.Journal of Chemical Informa- tion and Modeling2012,52, 2287–2299

    Desaphy, J.; Azdimousa, K.; Kellenberger, E.; Rog- nan, D. Comparison and druggability prediction of protein–ligand binding sites from pharmacophore- annotated cavity shapes.Journal of Chemical Informa- tion and Modeling2012,52, 2287–2299

  76. [84]

    How to measure the similarity between protein ligand-binding sites?Current Computer-Aided Drug Design2008,4, 209

    Kellenberger, E.; Schalon, C.; Rognan, D. How to measure the similarity between protein ligand-binding sites?Current Computer-Aided Drug Design2008,4, 209

  77. [85]

    Why is Tanimoto index an appropriate choice for fingerprint-based sim- ilarity calculations?Journal of Cheminformatics2015, 7, 1–13

    Bajusz, D.; R ´acz, A.; H ´eberger, K. Why is Tanimoto index an appropriate choice for fingerprint-based sim- ilarity calculations?Journal of Cheminformatics2015, 7, 1–13

  78. [86]

    https://github.com/ gnina/tfbio, 2018; Accessed: YYYY-MM-DD

    Garc ´ıa-Garc´ıa, J.; Others tfbio: Molecular features fea- turization library for TensorFlow. https://github.com/ gnina/tfbio, 2018; Accessed: YYYY-MM-DD

    2018

  79. [87]

    SoftSeg: Advan- tages of soft versus binary training for image segmen- tation.Medical Image Analysis2021,71, 102038

    Gros, C.; Lemay, A.; Cohen-Adad, J. SoftSeg: Advan- tages of soft versus binary training for image segmen- tation.Medical Image Analysis2021,71, 102038

  80. [88]

    Wang, Z.; Popordanoska, T.; Bertels, J.; Lemmens, R.; Blaschko, M. B. Dice semimetric losses: Optimizing the dice score with soft labels. International Conference on Medical Image Computing and Computer-Assisted Intervention. Cham, 2023; pp 475–485

    2023

  81. [89]

    V-net: Fully convolutional neural networks for volumetric medical image segmentation

    Milletari, F.; Navab, N.; Ahmadi, S.-A. V-net: Fully convolutional neural networks for volumetric medical image segmentation. 2016 Fourth International Confer- ence on 3D Vision (3DV). 2016; pp 565–571

    2016

  82. [90]

    Focal Loss for Dense Object Detection

    Lin, T.-Y .; Goyal, P.; Girshick, R.; He, K.; Doll ´ar, P. Focal Loss for Dense Object Detection. Proceedings of the IEEE International Conference on Computer Vision (ICCV). 2017; pp 2980–2988

    2017

  83. [91]

    Nair, V .; Hinton, G. E. Rectified Linear Units Improve Restricted Boltzmann Machines. Proceedings of the 27th International Conference on Machine Learning (ICML). 2010; pp 807–814

    2010

  84. [92]

    Pedregosa, F. et al. Scikit-learn: Machine Learning in Python.Journal of Machine Learning Research2011, 12, 2825–2830

  85. [93]

    Virtanen, P. et al. SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python.Nature Methods 2020,17, 261–272

    2020

  86. [94]

    Al- phaFold Protein Structure Database: massively expand- ing the structural coverage of protein-sequence space with high-accuracy models.Nucleic Acids Research 2022,50, D439–D444

    Varadi, M.; Anyango, S.; Deshpande, M.; et al. Al- phaFold Protein Structure Database: massively expand- ing the structural coverage of protein-sequence space with high-accuracy models.Nucleic Acids Research 2022,50, D439–D444

    2022

  87. [95]

    Rdkit documentation.Release2013,1, 4

    Landrum, G. Rdkit documentation.Release2013,1, 4

  88. [96]

    Comparative assessment of scoring functions: the CASF-2016 update.Journal of chemical information and modeling2018,59, 895–913

    Su, M.; Yang, Q.; Du, Y .; Feng, G.; Liu, Z.; Li, Y .; Wang, R. Comparative assessment of scoring functions: the CASF-2016 update.Journal of chemical information and modeling2018,59, 895–913

    2016

  89. [97]

    DELPHI Scorer: A Machine-Learning-Based Scoring Function for Protein- Ligand Interactions.Journal of Chemical Information and Modeling2017,57, 47–55

    Li, H.; Han, L.; Liu, W.; Wang, R. DELPHI Scorer: A Machine-Learning-Based Scoring Function for Protein- Ligand Interactions.Journal of Chemical Information and Modeling2017,57, 47–55

  90. [98]

    Boosting protein–ligand binding pose prediction and virtual screening based on residue–atom distance likelihood potential and graph transformer

    Shen, C.; Zhang, X.; Deng, Y .; Gao, J.; Wang, D.; Xu, L.; Kang, Y . Boosting protein–ligand binding pose prediction and virtual screening based on residue–atom distance likelihood potential and graph transformer. Journal of Medicinal Chemistry2022,65, 10691– 10706

  91. [99]

    L.; Kubra, B.; Sharaf, M.; Emwas, A.-H.; Jaremko, M.; Abdalla, M

    Faisal, S.; Badshah, S. L.; Kubra, B.; Sharaf, M.; Emwas, A.-H.; Jaremko, M.; Abdalla, M. Computa- tional study of SARS-CoV-2 rna dependent rna poly- merase allosteric site inhibition.Molecules2021,27, 223

  92. [100]

    Parvez, M. S. A.; Karim, M. A.; Hasan, M.; Jaman, J.; Karim, Z.; Tahsin, T.; Hasan, M. N.; Hosen, M. J. Prediction of potential inhibitors for RNA-dependent RNA polymerase of SARS-CoV-2 using comprehensive drug repurposing and molecular docking approach.In- ternational journal...

  93. [101]

    W.; Hadv ´ary, P

    Banner, D. W.; Hadv ´ary, P. Crystallographic analysis at 3.0-A resolution of the binding to human thrombin of four active site-directed inhibitors.Journal of Biological Chemistry1991,266, 20085–20093

  94. [102]

    I.; Buller, H

    Weitz, J. I.; Buller, H. R. Direct thrombin inhibitors in acute coronary syndromes: present and future.Circula- tion2002,105, 1004–1011

  95. [103]

    Crawley, J. T. B.; Zanardelli, S.; Chion, C. K. N. K.; Lane, D. A. The central role of thrombin in hemostasis. Journal of Thrombosis and Haemostasis2007,5, 95– 101

  96. [104]

    E.; Huntington, J

    Adams, T. E.; Huntington, J. A. Thrombin-cofactor interactions: structural insights into regulatory mecha- nisms.Arteriosclerosis, thrombosis, and vascular biol- ogy2006,26, 1738–1745

  97. [105]

    Exosites in the sub- strate specificity of blood coagulation reactions.Journal of Thrombosis and Haemostasis2007,5, 81–94

    Bock, P.; Panizzi, P.; Verhamme, I. Exosites in the sub- strate specificity of blood coagulation reactions.Journal of Thrombosis and Haemostasis2007,5, 81–94

  98. [106]

    Exosite binding in thrombin: a global struc- tural/dynamic overview of complexes with aptamers and other ligands.International Journal of Molecular Sciences2021,22, 10803

    Troisi, R.; Balasco, N.; Autiero, I.; Vitagliano, L.; Sica, F. Exosite binding in thrombin: a global struc- tural/dynamic overview of complexes with aptamers and other ligands.International Journal of Molecular Sciences2021,22, 10803

  99. [107]

    S.; Stafford, A

    Petrera, N. S.; Stafford, A. R.; Leslie, B. A.; Kretz, C. A.; Fredenburgh, J. C.; Weitz, J. I. Long range communication between exosites 1 and 2 modulates thrombin function.Journal of Biological Chemistry 2009,284, 25620–25629

    2009

  100. [108]

    Drug-protein interactions: refined structures of three sulfonamide drug complexes of human carbonic anhydrase I enzyme.Journal of molecular biology1994,243, 298–309

    Chakravarty, S.; Kannan, K. Drug-protein interactions: refined structures of three sulfonamide drug complexes of human carbonic anhydrase I enzyme.Journal of molecular biology1994,243, 298–309

  101. [109]

    H.; Robbins, A

    Sippel, K. H.; Robbins, A. H.; Domsic, J.; Genis, C.; Agbandje-McKenna, M.; McKenna, R. High-resolution structure of human carbonic anhydrase II complexed with acetazolamide reveals insights into inhibitor drug design.Acta Crystallographica Section F: Structural Biology and Cr...

  102. [110]

    Pfeiffer, N. Innovative glaucoma therapy: Glaucoma therapy with topical carbonic anhydrase inhibitors.Der Ophthalmologe: Zeitschrift der Deutschen Ophthalmol- ogischen Gesellschaft2001,98, 953–960

  103. [111]

    Supuran, C. T. Carbonic anhydrases: Novel therapeutic applications for inhibitors and activators.Nature Re- views Drug Discovery2008,7, 168–181

  104. [112]

    N.; Lindskog, S

    Silverman, D. N.; Lindskog, S. The catalytic mechanism of carbonic anhydrase: implications of a rate-limiting protolysis of water.Accounts of Chemical Research 1988,21, 30–36

    1988

  105. [113]

    K.; Christianson, D

    Nair, S. K.; Christianson, D. W. Unexpected pH- dependent conformation of His-64, the proton shuttle of carbonic anhydrase II.Journal of the American Chemical Society1991,113, 9455–9458

  106. [114]

    L.; Andring, J

    Lomelino, C. L.; Andring, J. T.; McKenna, R. Crys- tallography and its impact on carbonic anhydrase re- search.International journal of medicinal chemistry 2018,2018, 9419521

    2018

  107. [115]

    two-prong

    Jude, K. M.; Banerjee, A. L.; Haldar, M. K.; Manokaran, S.; Roy, B.; Mallik, S.; Srivastava, D.; Christianson, D. W. Ultrahigh resolution crystal struc- tures of human carbonic anhydrases I and II complexed with “two-prong” inhibitors reveal the molecular basis of high affinit...

  108. [116]

    M.; Banerjee, A

    Srivastava, D.; Jude, K. M.; Banerjee, A. L.; Hal- dar, M.; Manokaran, S.; Kooren, J.; Mallik, S.; Chris- tianson, D. W. Structural analysis of charge discrimi- nation in the binding of inhibitors to human carbonic anhydrases I and II.Journal of the American Chemical Society20...

  109. [117]

    Crystal structures of Lym- naea stagnalis AChBP in complex with neonicotinoid insecticides imidacloprid and clothianidin.Invertebrate Neuroscience2008,8, 71–81

    Ihara, M.; Okajima, T.; Yamashita, A.; Oda, T.; Hi- rata, K.; Nishiwaki, H.; Morimoto, T.; Akamatsu, M.; Ashikawa, Y .; Kuroda, S.; Mega, R.; Kuramitsu, S.; Sattelle, D.; Matsuda, K. Crystal structures of Lym- naea stagnalis AChBP in complex with neonicotinoid insecticides imi...

  110. [118]

    Nicotinic receptors: allosteric transi- tions and therapeutic targets in the nervous system

    Taly, A.; Corringer, P.-J.; Guedin, D.; Lestage, P.; Changeux, J.-P. Nicotinic receptors: allosteric transi- tions and therapeutic targets in the nervous system. Nature reviews Drug discovery2009,8, 733–750

  111. [119]

    P.; Holladay, M.; Williams, M

    Arneric, S. P.; Holladay, M.; Williams, M. Neuronal nicotinic receptors: a perspective on two decades of drug discovery research.Biochemical pharmacology 2007,74, 1092–1101

    2007

  112. [120]

    D.; Rezvani, A

    Levin, E. D.; Rezvani, A. H. Nicotinic interactions with antipsychotic drugs, models of schizophrenia and im- pacts on cognitive function.Biochemical Pharmacology 2007,74, 1182–1191

    2007

  113. [121]

    N.; Gratteri, P.; Guandalini, L.; Mar- tini, E.; Bonaccini, C.; Gualtieri, F

    Romanelli, M. N.; Gratteri, P.; Guandalini, L.; Mar- tini, E.; Bonaccini, C.; Gualtieri, F. Central nicotinic receptors: structure, function, ligands, and therapeutic potential.ChemMedChem: Chemistry Enabling Drug Discovery2007,2, 746–767

  114. [122]

    Cremer, J.; Le, T.; No ´e, F.; Clevert, D.-A.; Sch¨utt, K. T. PILOT: equivariant diffusion for pocket-conditioned de novo ligand generation with multi-objective guidance via importance sampling.Chemical Science2024,15, 14954–14967

  115. [123]

    A 3D pocket-aware and affinity-guided diffusion model for lead optimization.arXiv preprint arXiv:2504.21065 2025,

    Qiao, A.; Xie, J.; Huang, W.; Zhang, H.; Rao, J.; Zheng, S.; Yang, Y .; Wang, Z.; Li, G.-B.; Lei, J. A 3D pocket-aware and affinity-guided diffusion model for lead optimization.arXiv preprint arXiv:2504.21065 2025,

    2025 arXiv

  116. [124]

    Diffusion models in de novo drug design.Journal of Chemical Information and Modeling2024,64, 7238–7256

    Alakhdar, A.; Poczos, B.; Washburn, N. Diffusion models in de novo drug design.Journal of Chemical Information and Modeling2024,64, 7238–7256

  117. [125]

    Zhang, O.; Zhang, J.; Jin, J.; Zhang, X.; Hu, R.; Shen, C.; Cao, H.; Du, H.; Kang, Y .; Deng, Y .; others ResGen is a pocket-aware 3D molecular generation model based on parallel multiscale modelling.Nature Machine Intelligence2023,5, 1020–1030

  118. [126]

    Huang, L.; Xu, T.; Yu, Y .; Zhao, P.; Chen, X.; Han, J.; Xie, Z.; Li, H.; Zhong, W.; Wong, K.-C.; others A dual diffusion model enables 3D molecule generation and lead optimization based on target pockets.Nature Communications2024,15, 2657

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.