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
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 →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
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.
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
- 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.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
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)
- [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.
- [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.
- [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.
- [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)
- [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).
- [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.
- [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.
- [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
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
free parameters (8)
- protein sequence max length =
1000
- embedding dimension =
64
- GCN hidden layers =
3
- attention heads =
4
- dilation rate =
4
- dropout rate =
0.2
- learning rate, batch size, epochs =
5e-4, 512, 1000
- dynamic descriptor min-max normalization constants =
dataset-specific xmin, xmax per feature
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.
- domain assumption The MD simulations in ATLAS are accurate and representative for proteins in BindingDB.
- domain assumption Random five-fold split without target-disjointness is a valid evaluation protocol for generalization.
- domain assumption BindingDB experimental affinity measurements are reliable and comparable across labs.
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.
Reference graph
Works this paper leans on
-
[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]
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]
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]
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
-
[5]
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]
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
-
[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]
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
Show all 58 references
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
- [16]
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
- [31]
-
[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
-
[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
-
[34]
Nature 450(7172):964–
Henzler-Wildman K, Kern D (2007) Dynamic personalities of proteins. Nature 450(7172):964–
2007
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
- [45]
-
[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
-
[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
- [48]
-
[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
-
[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
-
[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
-
[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
-
[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
-
[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
- [55]
-
[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
-
[525]
https://doi.org/10.1002/pst.1943
1943 doi
-
[972]
https://doi.org/10.1038/nature06522
Reviewed August 15, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.