Pith. sign in

REVIEW 4 major objections 4 minor 58 references

DynamicDTA: Drug-Target Binding Affinity Prediction Using Dynamic Descriptors and Graph Representation

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Protein motion data sharpens drug-target affinity prediction.

desk verdict A sound architecture let down by an overstated abstract and an uncontrolled train/test split; the reported 'consistent' improvement is contradicted by the paper's own Table 2. read the letter →

arxiv 2505.11529 v1 pith:QLI6D4I5 submitted 2025-05-13 cs.RO cs.LG

Dan Luo , Jinyu Zhou , Le Xu , Sisi Yuan , Xuan Lin This is my paper · ORCID
classification cs.ROcs.LG
keywords drug-targetbindingaffinityproteindynamicsmoleculardescriptorsgraphneuralnetworkcross-attentiontensorfusionprediction
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

DynamicDTA proposes that drug-target binding affinity prediction is improved when a protein's dynamical behavior, not just its sequence or static structure, is given to the model. The paper's central claim is that adding four molecular-dynamics descriptors for each target—averaged residue fluctuation, radius of gyration, and two conformational-divergence measures—as an extra input modality yields consistently lower eRMSE and higher Pearson correlation than seven baselines on three affinity datasets. It reports gains of about 2.2% and 6.5% in eRMSE over the second-best model on the Kd and Ki benchmarks, and the best correlation on IC50 even though its eRMSE trails one image-based baseline slightly. If true, the result matters because it would mean cheap per-protein dynamics descriptors can supply information that sequence- and graph-only models miss, improving virtual screening without requiring target-ligand complex structures.

What carries the argument

The mechanism that carries the argument is a per-protein dynamic descriptor vector: four normalized numbers (average root mean square fluctuation, average radius of gyration, and two TM-score-derived divergence scores between simulation conformations) computed from molecular dynamics and attached to every target. The model pairs this vector with the drug's molecular graph embedding and the protein's dilated-convolution sequence embedding, lets the sequence and dynamics vectors attend to each other, and fuses all three through a tensor fusion network that computes their outer product. The descriptor vector is what distinguishes DynamicDTA from sequence- and graph-only baselines; the cross-attention and tensor fusion determine how that extra signal enters the prediction.

What would settle it

Re-run the same experiments with splits that never put the same protein in both training and test; if the eRMSE advantage over the best baselines on Kd* and Ki* falls below the reported 2-6% or disappears, the central claim of a dynamics-driven gain is falsified. A simpler check is to compare DynamicDTA against a variant that replaces the four descriptors with a learned per-protein embedding of the same size.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that a deep network taking a drug molecular graph, a protein sequence, and a four-dimensional vector of per-protein dynamics descriptors can beat seven existing predictors. The descriptors are treated as a constant property of each protein, normalized and passed through an MLP; a multi-head cross-attention layer lets the sequence embedding and the dynamics vector refine each other, and a tensor fusion network forms the outer product of the ligand, target, and dynamics embeddings so that unimodal, bimodal, and trimodal interactions are all represented before regression. Across the Kd*, Ki*, and IC50* datasets the model reports the best Pearson correlation in all three cases, the best eRMSE on Kd* and Ki*, and an eRMSE slightly above the best baseline on IC50*. On the held-out external set it reports roughly a 4.3% eRMSE improvement over the second-best method without fine-tuning, and ablations attribute part of the gain to the dynamics descriptors, especially the two TM-score-based divergence measures.

Load-bearing premise

The evaluation assumes that the five-fold cross-validation splits are protein-disjoint; if the same protein appears in both training and test, its per-protein dynamics vector leaks and can inflate the reported gains over baselines that lack such a feature.

Editorial extensions

If this is right

  • Adding MD-derived descriptors as a fourth modality reduces eRMSE by about 2.2% on Kd* and 6.5% on Ki* relative to the second-best baseline, with the largest gains concentrated in the more compact affinity distribution of Ki*.
  • Removing the dynamics descriptors degrades performance, and removing gyration radius plus the two divergence measures together produces the largest drop, so conformational-divergence descriptors carry most of the signal.
  • Tensor fusion outperforms concat, sum, average, and Hadamard fusion, indicating that higher-order interactions across ligand, sequence, and dynamics matter for the prediction.
  • A model trained on the three affinity datasets transfers to an unseen external set without fine-tuning, reporting roughly 4.3% lower eRMSE than the next-best method.
  • The cross-attention weights highlight residues in experimentally characterized binding pockets of the 2FOS complex, suggesting the dynamic signal is not only predictive but also interpretable.

Reading between the lines

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

  • Because the descriptor vector is constant per protein, an important control would be to replace it with a one-hot protein identity: if the gain persists, the model is learning protein-level affinity biases rather than a dynamics-based interaction signal.
  • If protein-disjoint splits erase the reported gains on Kd* and Ki*, the improvement would be attributable to descriptor leakage across training and test folds rather than to generalizable dynamics information.
  • The same architecture could be tested on per-residue or per-complex dynamic features, such as ligand-induced flexibility changes, instead of whole-protein constants; that would reveal whether the signal comes from binding-site plasticity.
  • The paper's own proposed extension of generating MD descriptors with diffusion models would turn an expensive per-protein preprocessing step into a fast prediction step, making the method practical for large-scale screening.
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 / 4 minor

Summary. The paper proposes DynamicDTA, a deep learning framework for drug-target binding affinity (DTA) prediction that combines three input modalities: a molecular graph representation of the drug processed by a graph convolutional network, the protein sequence encoded with dilated convolutions, and a four-dimensional vector of MD-derived dynamic descriptors (Avg.RMSF, Avg.Gyr, Div.SE, Div.MM) processed by an MLP. The protein sequence and dynamic vectors are fused with a multi-head cross-attention mechanism, and a tensor fusion network integrates all modalities before regression. The authors evaluate on three BindingDB-derived datasets (Kd*, Ki*, IC50*) and an external Kiba* dataset, comparing against seven baselines. The central claim is that DynamicDTA 'consistently outperforms all baseline methods' and achieves 'at least 3.4% improvement in eRMSE' across the three datasets; a case study on HIV-1 and interpretability analyses are also presented.

Significance. If the reported results were valid, the idea of injecting MD-derived protein flexibility descriptors as an additional input modality would be a useful contribution to DTA prediction, and the public availability of code and data is a strength. However, the evaluation evidence as presented is internally inconsistent with the headline claims, and the cross-validation protocol is described too loosely to rule out data leakage through per-protein dynamic descriptors. These issues directly undermine the central quantitative claim, so the current version does not establish the stated improvements.

major comments (4)
  1. [Abstract; Section 3.4; Table 2] The abstract and Section 3.4 claim that DynamicDTA 'achieves by at least 3.4% improvement in eRMSE' and 'consistently outperforms all baseline methods' on the three datasets. Table 2 directly contradicts this: on IC50*, DynamicDTA has eRMSE 0.611, worse than ImageDTA's 0.600, an 1.8% deficit; on Kd*, the improvement over GraphDTA is from 1.029 to 1.007, i.e., 2.1%, not 3.4%; only on Ki* does the improvement reach 8.2%. The Section 3.4 text even acknowledges the IC50* shortfall ('falls slightly behind ImageDTA in eRMSE'), which contradicts 'consistently outperforms all baseline methods.' The headline claim as stated is thus falsified by the paper's own results and must be corrected or substantially qualified.
  2. [Section 2.1; Section 2.2.1; Eq. (2)] The five-fold cross-validation splits are described only as 'each dataset was split into five parts,' without specifying whether the splits are disjoint at the protein or ligand level. Because the dynamic descriptor vector is a constant per protein (Section 2.2.1), a protein appearing in both training and test folds would leak its descriptor and enable the model to memorize protein-level affinity biases, while the baselines, which lack such per-protein constants, would not benefit equally. This makes the comparison potentially unfair. Additionally, the min-max normalization in Eq. (2) uses x_min and x_max computed 'across the dataset'; if computed over the full dataset including test folds, that is another form of leakage. The authors must clarify the split granularity, and ideally rerun the evaluation with a target-disjoint split to demonstrate generalization to unseen proteins.
  3. [Section 3.5.1; Table 3] The ablation study claims that replacing dilated convolutions with standard convolutions 'resulted in a decline in both eRMSE and R metrics.' However, on IC50*, the 'w/o Dilated' row reports eRMSE 0.611 and R 0.923, identical to the full DynamicDTA model; on Kd* and Ki*, the eRMSE differences are 0.014 and 0.005, respectively. These differences are well within the reported standard deviations (e.g., Kd* std of 0.050-0.053) and do not support the stated conclusion that dilated convolution is important. The text should be revised to acknowledge that the effect is negligible on IC50* and not statistically significant, or the claim should be supported with significance testing.
  4. [Section 3.7; Table 5] The Kiba* external validation is presented as evidence of 'better generalization ability,' but the absolute performance is weak: DynamicDTA achieves eRMSE 5.090 and R 0.202, and the second-best model has R 0.176. The claimed '13% improvement in R' is relative to an almost-zero baseline, and the eRMSE values are several-fold larger than on the main datasets, suggesting the model does not transfer well to a fully disjoint dataset. The statement 'further highlight the better generalization ability' is an overstatement given these numbers; at minimum, the authors should discuss why the absolute performance is so much worse and temper the generalization claim.
minor comments (4)
  1. [Throughout] There are numerous typographical errors that should be corrected: 'crated' (Section 2.1), 'benefical' (Abstract), 'aromaitic' (Section 2.2.1), 'enhancs' (Section 3.2), 'outperformes' (Section 3.4), and 'Liner Regression' (Table 5).
  2. [Section 3.8.2; Table 6] The HIV-1 case study lists five of ten top drugs as 'Unconfirmed' in PubMed evidence, and the docking scores reported in Fig. 6 (3.58-4.75) are not contextualized against any threshold or comparison, so the statement that these are 'relatively high' is not substantiated.
  3. [Section 3.1] The hyperparameter settings are described, but no information is given about how they were chosen; the parameter sensitivity analysis in Section 3.6 is performed only on Ki*, and it is unclear whether the same optimal values were used for the other datasets.
  4. [Section 3.4] The scatter plots in Figure 2 compare only DynamicDTA against DEAttentionDTA, not against the strongest baselines (ImageDTA on IC50* or GraphDTA on Kd*/Ki*), so the visual claim of 'tighter clustering' is not a complete comparison.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: dynamic descriptors and affinity targets are external data, and the derivation is self-contained.

full rationale

I find no circular derivation chain in this paper. The dynamic descriptors are external inputs taken from the ATLAS database (ref. 32), and the affinity regression targets are experimental measurements from BindingDB (ref. 41). The model is trained to map these inputs to those targets; nothing in the derivation fits a target-derived quantity and then renames it as a prediction. The HIV-1 case study is a post-training application used to rank drug candidates and is not used to fit any model parameter, so it is not an instance of fitted input being called a prediction. The only apparent author self-citation is ref. 9, a general overview of deep learning for drug repurposing co-authored by Xuan Lin; it supports a background statement about computational methods rather than the central performance claim, so it is not load-bearing. Concerns about protein-level split leakage and the apparent contradiction between Table 2 and the abstract's 'at least 3.4%' claim are evaluation-correctness risks, not circularity.

Assumptions & free parameters 8 free parameters · 4 assumptions · 0 invented entities

No new theoretical entities are introduced. The free parameters are model hyperparameters and normalization constants, none derived from first principles. The most important unstated burden is the evaluation split assumption, because dynamic descriptors are protein-specific constants and random sample-level splits can leak protein identity between train and test.

free parameters (8)
  • protein sequence max length = 1000
    Hand-chosen truncation threshold in Section 2.2.1; determines input representation.
  • embedding dimension = 64
    Hand-chosen in Section 3.1; used across all encoders.
  • GCN hidden layers = 3
    Tuned on the Ki* dataset in Section 3.6 (Figure 4d).
  • attention heads = 4
    Tuned on Ki* in Section 3.6 (Figure 4c).
  • dilation rate = 4
    Tuned on Ki* in Section 3.6 (Figure 4b).
  • dropout rate = 0.2
    Tuned on Ki* in Section 3.6 (Figure 4a).
  • learning rate, batch size, epochs = 5e-4, 512, 1000
    Hand-chosen in Section 3.1 with no sensitivity analysis on these values.
  • dynamic descriptor min-max normalization constants = dataset-specific xmin, xmax per feature
    Computed from the training data in Section 2.2.1 (Equation 2), making descriptors dataset-dependent.
assumptions (4)
  • domain assumption The four ATLAS MD descriptors (Avg.RMSF, Avg.Gyr, Div.SE, Div.MM) are sufficient to represent protein dynamics relevant to binding affinity.
    Used as the only dynamic input in Section 2.2.1; if these descriptors miss essential dynamics, the model cannot capture them.
  • domain assumption The MD simulations in ATLAS are accurate and representative for proteins in BindingDB.
    The model relies on ATLAS dynamic features matched by PDB ID in Section 2.1.
  • domain assumption Random five-fold split without target-disjointness is a valid evaluation protocol for generalization.
    Section 2.1 describes only a five-fold split and does not mention restricting proteins or drugs between train and test; this is load-bearing because dynamic descriptors are per-protein constants.
  • domain assumption BindingDB experimental affinity measurements are reliable and comparable across labs.
    The regression targets come from BindingDB with a negative log transform in Equation 1; the paper itself notes measurement inconsistencies as a limitation in the Discussion.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DynamicDTA: Drug-Target Binding Affinity Prediction Using Dynamic Descriptors and Graph Representation." pith.science (2026). https://pith.science/paper/QLI6D4I5

@misc{pith2026250511529,
  author       = {Pith},
  title        = {Pith review of: DynamicDTA: Drug-Target Binding Affinity Prediction Using Dynamic Descriptors and Graph Representation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QLI6D4I5}},
  note         = {Machine review of arXiv:2505.11529}
}
read the original abstract

Predicting drug-target binding affinity (DTA) is essential for identifying potential therapeutic candidates in drug discovery. However, most existing models rely heavily on static protein structures, often overlooking the dynamic nature of proteins, which is crucial for capturing conformational flexibility that will be beneficial for protein binding interactions. We introduce DynamicDTA, an innovative deep learning framework that incorporates static and dynamic protein features to enhance DTA prediction. The proposed DynamicDTA takes three types of inputs, including drug sequence, protein sequence, and dynamic descriptors. A molecular graph representation of the drug sequence is generated and subsequently processed through graph convolutional network, while the protein sequence is encoded using dilated convolutions. Dynamic descriptors, such as root mean square fluctuation, are processed through a multi-layer perceptron. These embedding features are fused with static protein features using cross-attention, and a tensor fusion network integrates all three modalities for DTA prediction. Extensive experiments on three datasets demonstrate that DynamicDTA achieves by at least 3.4% improvement in RMSE score with comparison to seven state-of-the-art baseline methods. Additionally, predicting novel drugs for Human Immunodeficiency Virus Type 1 and visualizing the docking complexes further demonstrates the reliability and biological relevance of DynamicDTA.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 33 canonical work pages

  1. [1]

    Nature Computational Sci- ence 3(12):1007–1008

    Le NQK (2023) Predicting emerging drug inter- actions using gnns. Nature Computational Sci- ence 3(12):1007–1008. https://doi.org/10.1038/ s43588-023-00555-7

    work page 2023

  2. [2]

    Computers in Biology and Medicine p 108435

    Zhang H, Liu X, Cheng W et al (2024) Pre- diction of drug-target binding affinity based on deep learning models. Computers in Biology and Medicine p 108435. https://doi.org/10.1016/ j.compbiomed.2024.108435

    arXiv 2024

  3. [3]

    Methods 231:1–7

    Tang X, Ma W, Yang M et al (2024) Mff-dta: Multi-scale feature fusion for drug-target affinity prediction. Methods 231:1–7. https://doi.org/10. 1016/j.ymeth.2024.08.008

    work page 2024

  4. [4]

    Interdisciplinary Sciences: 16 Computational Life Sciences 15(2):306–315

    D’Souza S, Prema K, Balaji S et al (2023) Deep learning-based modeling of drug–target interac- tion prediction incorporating binding site infor- mation of proteins. Interdisciplinary Sciences: 16 Computational Life Sciences 15(2):306–315. https://doi.org/10.1007/s12539-023-00557-z

    work page doi:10.1007/s12539-023-00557-z 2023

  5. [5]

    Nature 616(7958):673–685

    Sadybekov A V , Katritch V (2023) Computa- tional approaches streamlining drug discovery. Nature 616(7958):673–685. https://doi.org/10. 1038/s41586-023-05905-z

    work page 2023

  6. [6]

    Expert Systems with Applications 186:115810

    Tanoori B, Jahromi MZ, Mansoori EG (2021) Drug-target continuous binding affinity predic- tion using multiple sources of information. Expert Systems with Applications 186:115810. https://doi.org/10.1016/j.eswa.2021.115810

    work page arXiv 2021

  7. [7]

    Briefings in Bioinformat- ics 24(6):bbad386

    Pei Q, Wu L, Zhu J et al (2023) Break- ing the barriers of data scarcity in drug–target affinity prediction. Briefings in Bioinformat- ics 24(6):bbad386. https://doi.org/10.1093/bib/ bbad386

    doi:10.1093/bib/ 2023

  8. [8]

    Expert Systems with Applications 255:124647

    Zhu Z, Zheng X, Qi G et al (2024) Drug–target binding affinity prediction model based on multi- scale diffusion and interactive learning. Expert Systems with Applications 255:124647. https:// doi.org/10.1016/j.eswa.2024.124647

    work page arXiv 2024

Show all 58 references
  1. [9]

    Wiley interdisciplinary reviews: Computational molecular science 12(4):e1597

    Pan X, Lin X, Cao D et al (2022) Deep learning for drug repurposing: Methods, databases, and applications. Wiley interdisciplinary reviews: Computational molecular science 12(4):e1597. https://doi.org/10.1002/wcms.1597

    2022 doi

  2. [10]

    Interdisciplinary Sciences: Computational Life Sciences 15(4):696–709

    Zhang X, Li Y , Wang J et al (2023) A multi- perspective model for protein–ligand-binding affinity prediction. Interdisciplinary Sciences: Computational Life Sciences 15(4):696–709. https://doi.org/10.1007/s12539-023-00582-y

    2023 doi

  3. [11]

    Journal of Chemical Information and Modeling 54(3):735–743

    Tang J, Szwajda A, Shakyawar S et al (2014) Making sense of large-scale kinase inhibitor bioactivity data sets: a comparative and integra- tive analysis. Journal of Chemical Information and Modeling 54(3):735–743. https://doi.org/10. 1021/ci400709d

    2014

  4. [12]

    Nature biotechnology 29(11):1046–1051

    Davis MI, Hunt JP, Herrgard S et al (2011) Com- prehensive analysis of kinase inhibitor selec- tivity. Nature biotechnology 29(11):1046–1051. https://doi.org/10.1038/nbt.1990

    2011 doi

  5. [13]

    Medicine in Drug Discovery 17:100149

    V o TH, Nguyen NTK, Le NQK (2023) Improved prediction of drug-drug interactions using ensemble deep neural networks. Medicine in Drug Discovery 17:100149. https://doi.org/10. 1016/j.medidd.2022.100149

    2023

  6. [14]

    Bioinformat- ics 34(21):3666–3674

    Stepniewska-Dziubinska MM, Zielenkiewicz P, Siedlecki P (2018) Development and evalua- tion of a deep learning model for protein– ligand binding affinity prediction. Bioinformat- ics 34(21):3666–3674. https://doi.org/10.1093/ bioinformatics/bty374

    2018

  7. [15]

    Bioinformatics 34(17):i821–i829

    ¨Ozt¨urk H, ¨Ozg¨ur A, Ozkirimli E (2018) Deep- dta: deep drug–target binding affinity predic- tion. Bioinformatics 34(17):i821–i829. https:// doi.org/10.1093/bioinformatics/bty593

    2018 doi

  8. [16]

    arXiv preprint arXiv:190204166 https://doi.org/ 10.48550/arXiv.1902.04166

    ¨Ozt¨urk H, Ozkirimli E, ¨Ozg¨ur A (2019) Wid- edta: prediction of drug-target binding affinity. arXiv preprint arXiv:190204166 https://doi.org/ 10.48550/arXiv.1902.04166

    2019 doi arXiv

  9. [17]

    IEEE/ACM transactions on computational biology and bioinformatics 20(2):852–863

    Zhao Q, Duan G, Yang M et al (2022) Atten- tiondta: Drug–target binding affinity predic- tion by sequence-based deep learning with attention mechanism. IEEE/ACM transactions on computational biology and bioinformatics 20(2):852–863. https://doi.org/10.1109/TCBB. 2022.3170365

    2022

  10. [18]

    Bioinformatics p btae319

    Chen X, Huang J, Shen T et al (2024) Deat- tentiondta: Protein-ligand binding affinity pre- diction based on dynamic embedding and self- attention. Bioinformatics p btae319. https://doi. org/10.1093/bioinformatics/btae319

    2024 doi

  11. [19]

    Interdisciplinary Sciences: Computational Life Sciences pp 1–20

    Li G, Yuan Y , Zhang R (2024) Predict- ing protein–ligand binding affinity using fusion model of spatial-temporal graph neural network and 3d structure-based complex graph. Interdisciplinary Sciences: Computational Life Sciences pp 1–20. https://doi.org/10.1007/s12539-024-00644-9

    2024 doi

  12. [20]

    Bioinformat- ics 37(8):1140–1147

    Nguyen T, Le H, Quinn TP et al (2021) Graphdta: predicting drug–target binding affin- ity with graph neural networks. Bioinformat- ics 37(8):1140–1147. https://doi.org/10.1093/ bioinformatics/btaa921

    2021

  13. [21]

    RSC advances 10(35):20701– 20712

    Jiang M, Li Z, Zhang S et al (2020) Drug–target affinity prediction using graph neural network 17 and contact maps. RSC advances 10(35):20701– 20712. https://doi.org/10.1039/D0RA02297G

    2020 doi

  14. [22]

    ACS omega 9(26):28485–28493

    Han L, Kang L, Guo Q (2024) Imagedta: A sim- ple model for drug–target binding affinity pre- diction. ACS omega 9(26):28485–28493. https: //doi.org/10.1021/acsomega.4c02308

    2024 doi

  15. [23]

    nature 596(7873):583–589

    Jumper J, Evans R, Pritzel A et al (2021) Highly accurate protein structure prediction with alphafold. nature 596(7873):583–589. https:// doi.org/10.1038/s41586-021-03819-2

    2021 doi

  16. [24]

    Nucleic acids research 47(D1):D520–D528

    (2019) Protein data bank: the single global archive for 3d macromolecular structure data. Nucleic acids research 47(D1):D520–D528. https://doi.org/10.1093/nar/gky949

    2019 doi

  17. [25]

    Information Sciences 613:507–523

    Chu Z, Huang F, Fu H et al (2022) Hierarchi- cal graph representation learning for the predic- tion of drug-target binding affinity. Information Sciences 613:507–523. https://doi.org/10.1016/j. ins.2022.09.043

    2022 doi

  18. [26]

    Interdisciplinary Sciences: Computa- tional Life Sciences pp 1–11

    Wang S, Liu Y , Zhang Y et al (2024) Chl-dti: A novel high–low order information convergence framework for effective drug–target interaction prediction. Interdisciplinary Sciences: Computa- tional Life Sciences pp 1–11. https://doi.org/10. 1007/s12539-024-00608-z

    2024

  19. [27]

    Nature Communications 15(1):1071

    Lu W, Zhang J, Huang W et al (2024) Dynamicbind: Predicting ligand-specific protein-ligand complex structure with a deep equivariant generative model. Nature Communications 15(1):1071. https://doi.org/10.1038/s41467-024-45461-2

    2024 doi

  20. [28]

    Brief- ings in Bioinformatics 25(2):bbae081

    Wang H (2024) Prediction of protein–ligand binding affinity via deep learning models. Brief- ings in Bioinformatics 25(2):bbae081. https:// doi.org/10.1093/bib/bbae081

    2024 doi

  21. [29]

    Interdisciplinary Sciences: Computational Life Sciences 11:115–124

    Hashemzadeh S, Ramezani F, Rafii-Tabar H (2019) Study of molecular mechanism of the interaction between mek1/2 and trametinib with docking and molecular dynamic simulation. Interdisciplinary Sciences: Computational Life Sciences 11:115–124. https://doi.org/10.1007/ s12539-018-0305-4

    2019

  22. [30]

    arXiv preprint arXiv:240812413 https://doi.org/ 10.48550/arXiv.2408.12413

    Liu C, Wang J, Cai Z et al (2024) Dynamic pdb: A new dataset and a se (3) model extension by integrating dynamic behaviors and physical properties in protein structures. arXiv preprint arXiv:240812413 https://doi.org/ 10.48550/arXiv.2408.12413

    2024 doi arXiv

  23. [31]

    arXiv preprint arXiv:170707250 https://doi

    Zadeh A, Chen M, Poria S et al (2017) Tensor fusion network for multimodal sentiment analy- sis. arXiv preprint arXiv:170707250 https://doi. org/10.48550/arXiv.1707.07250

    2017 doi arXiv

  24. [32]

    Nucleic acids research 52(D1):D384– D392

    Vander Meersche Y , Cretin G, Gheeraert A et al (2024) Atlas: protein flexibility descrip- tion from atomistic molecular dynamics sim- ulations. Nucleic acids research 52(D1):D384– D392. https://doi.org/10.1093/nar/gkad1084

    2024 doi

  25. [33]

    Nature Communications 15(1):5538

    Song X, Bao L, Feng C et al (2024) Accu- rate prediction of protein structural flexibility by deep learning integrating intricate atomic struc- tures and cryo-em density information. Nature Communications 15(1):5538. https://doi.org/10. 1038/s41467-024-49858-x

    2024

  26. [34]

    Nature 450(7172):964–

    Henzler-Wildman K, Kern D (2007) Dynamic personalities of proteins. Nature 450(7172):964–

    2007

  27. [35]

    Journal of molecu- lar liquids 354:118901

    Ghahremanian S, Rashidi MM, Raeisi K et al (2022) Molecular dynamics simulation approach for discovering potential inhibitors against sars- cov-2: A structural review. Journal of molecu- lar liquids 354:118901. https://doi.org/10.1016/j. molliq.2022.118901

    2022

  28. [36]

    Molecu- lar Biology 42:623–628

    Lobanov MY , Bogatyreva N, Galzitskaya O (2008) Radius of gyration as an indica- tor of protein structure compactness. Molecu- lar Biology 42:623–628. https://doi.org/10.1134/ S0026893308040195

    2008

  29. [37]

    Nucleic acids research 33(7):2302–2309

    Zhang Y , Skolnick J (2005) Tm-align: a protein structure alignment algorithm based on the tm- score. Nucleic acids research 33(7):2302–2309. https://doi.org/10.1093/nar/gki524

    2005 doi

  30. [38]

    Science 334(6055):517–520

    Lindorff-Larsen K, Piana S, Dror RO et al (2011) How fast-folding pro- teins fold. Science 334(6055):517–520. https://doi.org/10.1126/science.1208351 18

    2011 doi

  31. [39]

    Proteins: Structure, Function, and Bioin- formatics 80(8):2071–2079

    Raval A, Piana S, Eastwood MP et al (2012) Refinement of protein structure homology mod- els via long, all-atom molecular dynamics simu- lations. Proteins: Structure, Function, and Bioin- formatics 80(8):2071–2079. https://doi.org/10. 1002/prot.24098

    2012

  32. [40]

    Proceedings of the National Academy of Sciences 109(22):E1428–E1436

    Rivalta I, Sultan MM, Lee NS et al (2012) Allosteric pathways in imidazole glycerol phos- phate synthase. Proceedings of the National Academy of Sciences 109(22):E1428–E1436. https://doi.org/10.1073/pnas.1120536109

    2012 doi

  33. [41]

    Nucleic acids research 35(suppl 1):D198– D201

    Liu T, Lin Y , Wen X et al (2007) Bind- ingdb: a web-accessible database of experimen- tally determined protein–ligand binding affini- ties. Nucleic acids research 35(suppl 1):D198– D201. https://doi.org/10.1093/nar/gkl999

    2007 doi

  34. [42]

    Bioinformatics 39(2):btad056

    Hua Y , Song X, Feng Z et al (2023) Mfr-dta: a multi-functional and robust model for pre- dicting drug–target binding affinity and region. Bioinformatics 39(2):btad056. https://doi.org/ 10.1093/bioinformatics/btad056

    2023 doi

  35. [43]

    Weininger D (1988) Smiles, a chemical lan- guage and information system. 1. introduc- tion to methodology and encoding rules. Jour- nal of chemical information and computer sciences 28(1):31–36. https://doi.org/10.1021/ ci00057a005

    1988

  36. [44]

    Journal of Chemin- formatics 12:1–16

    Bento AP, Hersey A, F ´elix E et al (2020) An open source chemical structure curation pipeline using rdkit. Journal of Chemin- formatics 12:1–16. https://doi.org/10.1186/ s13321-020-00456-1

    2020

  37. [45]

    arXiv preprint arXiv:151107122 https://doi.org/10.48550/arXiv.1511.07122

    Yu F, Koltun V (2015) Multi-scale context aggregation by dilated convolu- tions. arXiv preprint arXiv:151107122 https://doi.org/10.48550/arXiv.1511.07122

    2015 doi arXiv

  38. [46]

    Bioinfor- matics 39(2):btad049

    Jin Z, Wu T, Chen T et al (2023) Capla: improved prediction of protein–ligand bind- ing affinity by a deep learning approach based on a cross-attention mechanism. Bioinfor- matics 39(2):btad049. https://doi.org/10.1093/ bioinformatics/btad049

    2023

  39. [47]

    Briefings in Bioinfor- matics 23(6):bbac468

    Zhang L, Wang CC, Chen X (2022) Predicting drug–target binding affinity through molecule representation block based on multi-head atten- tion and skip connection. Briefings in Bioinfor- matics 23(6):bbac468. https://doi.org/10.1093/ bib/bbac468

    2022

  40. [48]

    arXiv preprint arXiv:14126980 https://doi.org/10.48550/arXiv

    Kingma DP, Ba J (2014) Adam: A method for stochastic optimization. arXiv preprint arXiv:14126980 https://doi.org/10.48550/arXiv. 1412.6980

    2014 doi arXiv

  41. [49]

    Pharmaceutical Statistics 18(5):513–

    Barbet J, Huclier-Markai S (2019) Equilibrium, affinity, dissociation constants, ic5o: Facts and fantasies. Pharmaceutical Statistics 18(5):513–

    2019

  42. [50]

    Fron- tiers in Pharmacology 9:1136

    Lagunin AA, Romanova MA, Zadorozhny AD et al (2018) Comparison of quantitative and qual- itative (q) sar models created for the prediction of ki and ic50 values of antitarget inhibitors. Fron- tiers in Pharmacology 9:1136. https://doi.org/10. 3389/fphar.2018.01136

    2018

  43. [51]

    Bioinformatics 39(6):btad355

    He H, Chen G, Chen CYC (2023) Nhgnn-dta: a node-adaptive hybrid graph neural network for interpretable drug–target binding affinity predic- tion. Bioinformatics 39(6):btad355. https://doi. org/10.1093/bioinformatics/btad355

    2023 doi

  44. [52]

    Interdisciplinary Sciences: Computa- tional Life Sciences 11:320–328

    Li J, Fu A, Zhang L (2019) An overview of scoring functions used for protein– ligand interactions in molecular docking. Interdisciplinary Sciences: Computa- tional Life Sciences 11:320–328. https: //doi.org/10.1007/s12539-019-00327-w

    2019 doi

  45. [53]

    Bioinformatics p btae745

    ˇSkrh´ak V , Novotn`y M, Feidakis CP et al (2024) Cryptobench: Cryptic protein-ligand binding sites dataset and benchmark. Bioinformatics p btae745. https://doi.org/10.1093/bioinformatics/ btae745

    2024 doi

  46. [54]

    Nature Communica- tions 14(1):1177–1177

    Meller A, Ward M, Borowsky J et al (2023) Predicting locations of cryptic pockets from sin- gle protein structures using the pocketminer graph neural network. Nature Communica- tions 14(1):1177–1177. https://doi.org/10.1038/ s41467-023-36699-3

    2023

  47. [55]

    arXiv preprint arXiv:221001776 https: //doi.org/10.48550/arXiv.2210.01776 19

    Corso G, St ¨ark H, Jing B et al (2022) Diffdock: Diffusion steps, twists, and turns for molecular docking. arXiv preprint arXiv:221001776 https: //doi.org/10.48550/arXiv.2210.01776 19

    2022 doi arXiv

  48. [56]

    Nature Computational Science 4(12):899–909

    Schneuing A, Harris C, Du Y et al (2024) Structure-based drug design with equivari- ant diffusion models. Nature Computational Science 4(12):899–909. https://doi.org/10.1038/ s43588-024-00737-x 20

    2024

  49. [525]

    https://doi.org/10.1002/pst.1943

    1943 doi

  50. [972]

    https://doi.org/10.1038/nature06522

    doi

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.