Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

A 3D pocket-aware and affinity-guided diffusion model for lead optimization

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read Diffleop's affinity-guided denoising produces lead molecules that beat prior 3D generative models on predicted binding affinity.

desk verdict Diffleop is a sensible combination of known diffusion ingredients with a clear affinity-guidance story, but the evaluation is too thin (one run, proxy predictors only) to back the SOTA and 'enhanced affinity' headline. read the letter →

arxiv 2504.21065 v1 pith:2NDOBZWL submitted 2025-04-29 cs.LG cs.AI

Anjie Qiao , Junjie Xie , Weifeng Huang , Hao Zhang , Jiahua Rao , Shuangjia Zheng , Yuedong Yang , Zhen Wang
show 2 more authors
Guo-Bo Li Jinping Lei
This is my paper · ORCID
classification cs.LGcs.AI
keywords leadoptimization3Ddiffusionmodelbindingaffinityguidanceequivariantgraphneuralnetworkscaffolddecorationlinkerdesignstructure-baseddrugmoleculargeneration
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

Diffleop sets out to show that lead optimization in 3D—decorating a scaffold or linking fragments inside a protein pocket—can be driven explicitly by predicted binding affinity, not just by protein structure. It couples an equivariant diffusion model that generates atom coordinates, atom types, and bond types with a trained affinity predictor whose gradients steer each denoising step toward higher affinity. If the central claim holds, a generative model can take an existing lead and produce optimized candidates whose top-five predicted affinity matches or exceeds the reference ligand while keeping drug-like properties. The paper reports such results on two tasks, with high-affinity hit rates near 53 percent, far above the baselines it compares against. Automated, affinity-directed optimization of this kind is the step where computational drug design has historically been weakest.

What carries the argument

The load-bearing mechanism is a trained affinity predictor used as a guidance field around an E(3)-equivariant diffusion denoiser. From the denoiser's graph embeddings the predictor outputs $\hat A = \frac{1}{N_L}\sum_i \mathrm{sigmoid}(\mathrm{MLP}(L_t,P,t))$; at each reverse step the model shifts coordinates along $\nabla_{x_t}\hat A$ and multiplies one-hot atom and bond type vectors by $e^{r_1\nabla_{v_t}\hat A}$ and $e^{r_2\nabla_{b_t}\hat A}$. The denoiser itself is an equivariant graph neural network with a protein–ligand $k$-nearest-neighbor graph and a fully connected ligand graph carrying bond-type messages; fake atom and fake bond types make molecular size and connectivity decisions part of the diffusion process. The paper credits this combined machinery for both the affinity improvement (guidance) and the realistic drug-like properties (bond diffusion).

What would settle it

Run the same benchmark with guidance gradients replaced by random vectors of the same scale; if the random-guidance version matches Diffleop's affinity, the central claim is falsified. A more direct test: on held-out complexes, measure the rank correlation between the Equation (22) coordinate shift and the change in TANKBind affinity after completing generation; near-zero correlation would show the guidance signal carries no affinity information.

Watch

Extended reading notes

Core claim

On the paper's own terms, Diffleop is a pocket-aware diffusion model that treats bond types as diffused variables and injects binding affinity directly into sampling. An E(3)-equivariant predictor trained on noised protein-ligand complexes estimates $\hat A$, and during reverse diffusion its gradients shift atom coordinates ($\nabla_{x_t}\hat A$) and rescale one-hot atom and bond vectors by $e^{r_1\nabla_{v_t}\hat A}$ and $e^{r_2\nabla_{b_t}\hat A}$. In scaffold decoration across 100 pockets, the top 5 of 100 generated molecules reach average TANKBind-predicted affinity 6.71 versus 6.69 for the reference; in linker design, 6.74 versus 6.73. The high-affinity rate—the fraction of top-5 molecules meeting or beating the reference affinity—is 52.6% and 54.6%, compared with values at or below roughly 29% for the baselines. The paper concludes that affinity guidance is responsible for the affinity gain and that diffusing bond types improves drug-likeness.

Load-bearing premise

The whole result rests on the assumption that the small affinity predictor, trained on noised protein–ligand complexes, produces gradients that point toward genuinely higher binding affinity when applied to new scaffolds and pockets; the paper offers no direct test of that gradient quality, and the supporting ablation is a single run without error bars.

Editorial extensions

If this is right

  • If the central claim is correct, lead optimization becomes a closed 3D loop: fixed pocket plus retained scaffold or fragments is enough to generate candidates whose top-five predicted affinity is at or above the reference, with no separate bond-inference post-processing step.
  • Because the same affinity-guidance scheme is used for both scaffold decoration and linker design, one trained model can switch between the two lead-optimization protocols without retraining.
  • The reported high-affinity rates mean that, per 100 generated molecules, a user can typically find at least two to three candidates that match or beat the reference on predicted affinity, a usable hit rate for ranking and synthesis.
  • Joint diffusion of bond types yields complete molecules directly, and the reported QED, SA, LogP, and Lipinski values stay close to reference compounds, so affinity gains do not obviously come at the cost of synthetic accessibility or drug-likeness.

Reading between the lines

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

  • An implication the paper leaves implicit is that the guidance scales $s$, $r_1$, and $r_2$ form a continuous trade-off knob: larger gradients should push harder toward affinity but can distort atom and bond distributions, so the method is better viewed as a tunable sampler than a fixed generator.
  • The guidance recipe only requires a differentiable predictor, so the same machinery could be pointed at selectivity, toxicity, or ADMET endpoints, or even at a weighted sum of objectives; this is an extension the paper does not make.
  • If the affinity predictor is the real engine of the result, then its accuracy on noised complexes, not the diffusion architecture, likely sets the ceiling for achievable affinity; benchmarking that predictor independently would predict how well the method transfers to new target families.
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 / 5 minor

Summary. The manuscript presents Diffleop, a 3D pocket-aware diffusion model for lead optimization that generates atom positions, atom types, and bond types simultaneously inside a fixed protein pocket, conditioned on retained scaffold or fragments. The key novelty is affinity guidance: an E(3)-equivariant network is trained to predict binding affinity from noised protein-ligand complexes, and gradients of that predictor with respect to coordinates, atom types, and bond types steer the reverse diffusion process toward higher predicted affinity. The authors evaluate Diffleop on scaffold decoration and linker design tasks, comparing with GraphBP, AR, Pocket2Mol, and DiffLinker using TANKBind-predicted affinity, QED, SA, LogP, and Lipinski compliance, and report higher mean affinity and high-affinity percentage than baselines. Ablations indicate reduced affinity without affinity guidance and reduced drug-likeness without bond diffusion.

Significance. If the affinity-guidance mechanism is effective, Diffleop addresses a real limitation of prior 3D generative models that lack an explicit affinity-driven objective, and the joint bond diffusion is a useful methodological step. The manuscript ships no code or data, but its main value would be a reusable recipe for gradient-guided lead optimization. The significance is currently limited by the lack of validation of the guidance predictor and by the single-run evaluation; the claims should be read as preliminary rather than established.

major comments (4)
  1. [Section 3.5, Eqs. (21)-(24) and Table 3] The load-bearing assumption of the affinity guidance is that gradients of the trained predictor  with respect to noised atom coordinates, atom types, and bond types point toward molecules with higher TANKBind-predicted affinity on unseen scaffolds and pockets. The manuscript provides no train/test split for the affinity predictor, no report of its accuracy or calibration on noised inputs, and no correlation between  and TANKBind. The only direct evidence is the single-run ablation in Table 3, where without affinity guidance Affinity drops from 6.71 to 6.49 and High Affinity from 52.6% to 40.4%; with no error bars or multiple seeds, this gap is not distinguishable from sampling noise. I recommend reporting predictor validation and multi-seed error bars, and a per-timestep comparison of guided versus unguided TANKBind scores.
  2. [Section 3.5, Eqs. (23)-(24)] The multiplicative update for atom and bond types is not specified as operating on a probability vector, a one-hot vector, or logits, and no normalization step is described after the update. Starting from a categorical distribution, the output of vt←(vt+δ)·exp(r1∇vtÂ) need not be a valid probability distribution, which makes the exact generative process undefined. Please specify the domain of vt, the normalization or projection step, and the values of δ, r1, r2, s, λ1, λ2, T, and k.
  3. [Section 2.2, Tables 1-3] All quantitative claims of superiority are based on a single run with no error bars, no multiple seeds, and no statistical tests. For example, Table 1 reports Diffleop Affinity 6.71 versus Reference 6.69, a difference of 0.02, which is tiny relative to the granularity of TANKBind scores; and the High Affinity differences across ablations are large but without confidence intervals. The word 'significantly' is therefore not supported. Please add standard deviations over seeds and a paired significance test for the main comparisons.
  4. [Section 2.2 and Section 3.7] Affinity and High Affinity are computed with TANKBind, a machine-learned surrogate, and the guidance predictor is another machine-learned surrogate of the same type. The paper does not independently validate improved binding affinity with docking or experimental data. At minimum, show that the improvement transfers to an independent scoring function or docking program (e.g., Vina) and discuss the known error of TANKBind on generated molecules.
minor comments (5)
  1. [Section 3.1] The notation for retained groups uses N_L for the size of R ('i,j∈{1,...,N_L}'), which appears to be a typo for N_R; please correct.
  2. [Section 3.2] The recurrence relation \bar\sigma_t = \sigma_t - \bar\alpha_t \bar\sigma_{t-1} is not consistent with standard variance-preserving noising; please clarify the noise schedule and how \alpha_t and \sigma_t are defined.
  3. [Section 2.2] The text says Diffleop outperforms 'all' baselines, but the baseline sets differ between tasks; specify which baselines are compared for each task.
  4. [Section 3.5] The affinity predictor training loss and the data split used to train it are not described; please add these details.
  5. [Section 2.2] The test set is described only as '100 protein targets' with no dataset names or splits; please specify the source of complexes and the scaffold/fragment construction procedure.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the affinity-guidance objective (Eq. 21) and the TANKBind evaluation metric are distinct fitted functions, and the cited self-work is not load-bearing.

full rationale

Diffleop's derivation is self-contained rather than circular. The conditional diffusion model is trained with the standard loss terms in Eqs. 9-12, and binding affinity is injected at sampling time as classifier-style guidance from a separately fitted EGNN-MLP predictor (Eq. 21), whose gradients are applied to coordinates, atom types, and bond types in Eqs. 22-24. The paper's headline binding-affinity results, however, are computed with TANKBind on the final generated molecules, as stated in Section 2.2 and Section 3.7. Nothing in the manuscript defines the guidance predictor to be TANKBind, and no equation makes the predicted affinity A-hat equal to the TANKBind score by construction. The empirical question of whether gradients of the fitted predictor transfer to TANKBind on unseen scaffolds is a robustness or correctness risk (the paper does not report predictor calibration, gradient alignment, or error bars on the ablation), but it is not circularity. The only self-citations are to DiffDec for the fake-atom / fake-bond diffusion mechanism and as an example of prior diffusion-based scaffold decoration; those are design borrowings and are not load-bearing for the central affinity-guidance claim. No uniqueness theorem is imported, no fitted parameter is renamed as a prediction, and no known empirical pattern is repackaged as a derivation.

Assumptions & free parameters 6 free parameters · 5 assumptions · 2 invented entities

The central claim depends on several hyperparameters (gradient scales and loss weights) whose values are not reported, and on the empirical assumption that a learned affinity predictor trained on noised complexes guides generation toward higher TANKBind affinity on new targets. The fake atom and fake bond are invented modeling tokens with no physical meaning. The axioms are standard diffusion modeling assumptions plus domain assumptions about fixed pockets and TANKBind as an affinity proxy.

free parameters (6)
  • s (atom coordinate gradient scale)
    Introduced in Eq. 22 to scale the affinity gradient on atom coordinates; no value or tuning procedure is reported.
  • r1 (atom type gradient scale)
    Introduced in Eq. 23 to scale the affinity gradient on atom-type one-hot vectors; no value reported.
  • r2 (bond type gradient scale)
    Introduced in Eq. 24 to scale the affinity gradient on bond-type one-hot vectors; no value reported.
  • lambda1, lambda2 (loss weights)
    Predefined constants in Eq. 9 weighting atom-type and bond-type losses; values not reported.
  • T (number of diffusion steps)
    Diffusion horizon, not reported.
  • k (nearest neighbors in protein-ligand graph)
    kNN graph size in Eq. 13; value not reported.
assumptions (5)
  • standard math The forward diffusion processes in Eqs. 1-7 correctly define the noised distributions.
    These are standard Gaussian and categorical noising processes from the diffusion literature; they are stated but not proven.
  • standard math The reverse denoising transition p_theta(M_{t-1} | M_t, R, P) is well-approximated by the EGNN parameterization.
    Assumed by all score and diffusion models; standard approximation.
  • ad hoc to paper The affinity predictor in Eq. 21, trained on noised complexes, provides useful gradients for increasing TANKBind-predicted affinity on unseen targets.
    This is the key empirical assumption for the affinity-guidance mechanism; the paper provides only a single-run ablation as support.
  • domain assumption TANKBind-predicted binding affinity is a valid proxy for real binding affinity.
    All affinity evaluations use TANKBind predictions; the paper does not use any experimental affinity measurements.
  • domain assumption The protein pocket structures P are fixed and correctly represent the binding site throughout the diffusion process.
    Standard structure-based drug design assumption; pockets are treated as rigid.
invented entities (2)
  • Fake atom (mask padding)
    purpose: Integrates molecule size prediction into the diffusion model, following DiffDec (Xie et al. 2024).
    A modeling construct with no physical counterpart and no external evidence.
  • Fake bond type (bond padding)
    purpose: Enables diffusion on fully connected molecular graphs and eliminates post-processing bond inference.
    A modeling construct added to the bond type vocabulary; no physical counterpart and no external evidence.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A 3D pocket-aware and affinity-guided diffusion model for lead optimization." pith.science (2026). https://pith.science/paper/2NDOBZWL

@misc{pith2026250421065,
  author       = {Pith},
  title        = {Pith review of: A 3D pocket-aware and affinity-guided diffusion model for lead optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2NDOBZWL}},
  note         = {Machine review of arXiv:2504.21065}
}
read the original abstract

Molecular optimization, aimed at improving binding affinity or other molecular properties, is a crucial task in drug discovery that often relies on the expertise of medicinal chemists. Recently, deep learning-based 3D generative models showed promise in enhancing the efficiency of molecular optimization. However, these models often struggle to adequately consider binding affinities with protein targets during lead optimization. Herein, we propose a 3D pocket-aware and affinity-guided diffusion model, named Diffleop, to optimize molecules with enhanced binding affinity. The model explicitly incorporates the knowledge of protein-ligand binding affinity to guide the denoising sampling for molecule generation with high affinity. The comprehensive evaluations indicated that Diffleop outperforms baseline models across multiple metrics, especially in terms of binding affinity.

Figures

Figures reproduced from arXiv: 2504.21065 by the authors.

Figure 1
Figure 1. The architecture of Diffleop. (a) The framework of Diffleop. The diffusion process gradually adds predefined [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Diffleop achieves state-of-art performance on molecular optimization through scaffold decoration and linker [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

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

    q-bio.BM 2025-02 conditional novelty 4.0 of 10

    A pocket-prediction network with soft labels, ReLU output, and a five-model ensemble guides AutoDock Vina to 54.9% PoseBusters-valid top poses, beating DiffBindFR and approaching AlphaFold 3 on a time-split benchmark.

Reference graph

Works this paper leans on

52 extracted references · 38 canonical work pages · cited by 1 Pith paper

  1. [1]

    Coma: efficient structure-constrained molecular generation using contractive and margin losses

    Jinhyung Choi, Seungmin Seo, and Sung Ju Park. Coma: efficient structure-constrained molecular generation using contractive and margin losses. Journal of Cheminformatics, 15 0 (8), 2023

    work page 2023

  2. [2]

    Molecular optimization by capturing chemist's intuition using deep neural networks

    Jinlong He et al. Molecular optimization by capturing chemist's intuition using deep neural networks. Journal of Cheminformatics, 13: 0 26, 2021

    work page 2021

  3. [3]

    Efficient multi-objective molecular optimization in a continuous latent space

    Robin Winter et al. Efficient multi-objective molecular optimization in a continuous latent space. Chemical Science, 10: 0 8016--8024, 2019

    work page 2019

  4. [4]

    Min, Srinivasan Parthasarathy, and Xia Ning

    Zhenpeng Chen, Mingyang R. Min, Srinivasan Parthasarathy, and Xia Ning. A deep generative model for molecule optimization via one fragment modification. Nature Machine Intelligence, 3: 0 1040--1049, 2021

    work page 2021

  5. [5]

    Molecular de-novo design through deep reinforcement learning

    Marcus Olivecrona, Thomas Blaschke, Ola Engkvist, and Hongming Chen. Molecular de-novo design through deep reinforcement learning. Journal of Cheminformatics, 9: 0 48, 2017

    work page 2017

  6. [6]

    Fragrep: A web server for structure-based drug design by fragment replacement

    Jianjun Shan et al. Fragrep: A web server for structure-based drug design by fragment replacement. Journal of Chemical Information and Modeling, 60: 0 5900--5906, 2020

    work page 2020

  7. [7]

    de Souza Neto et al

    Luis R. de Souza Neto et al. In silico strategies to support fragment-to-lead optimization in drug discovery. Frontiers in Chemistry, 8: 0 93, 2020

    work page 2020

  8. [8]

    Target-focused library design by pocket-applied computer vision and fragment deep generative linking

    Mohamed Eguida et al. Target-focused library design by pocket-applied computer vision and fragment deep generative linking. Journal of Medicinal Chemistry, 65: 0 13771--13783, 2022

    work page 2022

Show all 52 references
  1. [9]

    Bos et al

    Pieter H. Bos et al. Autodesigner, a de novo design algorithm for rapidly exploring large chemical space for lead optimization: Application to the design and synthesis of d-amino acid oxidase inhibitors. Journal of Chemical Information and Modeling, 62: 0 1905--1915, 2022

    1905

  2. [10]

    Hoffman et al

    Samuel C. Hoffman et al. Optimizing molecules using efficient queries from property evaluations. Nature Machine Intelligence, 4: 0 21--31, 2021

    2021

  3. [11]

    Equivariant 3d-conditional diffusion model for molecular linker design

    Ilia Igashov, Hannes Stärk, Clément Vignac, Arne Schneuing, Victor Garcia Satorras, Pascal Frossard, Max Welling, Michael Bronstein, and Bruno Correia. Equivariant 3d-conditional diffusion model for molecular linker design. Nature Machine Intelligence, 6: 0 417--427, 2024. doi...

    2024 doi

  4. [12]

    J. Wang, S. Zheng, J. Chen, and Y. Yang. Meta learning for low-resource molecular optimization. Journal of Chemical Information and Modeling, 61 0 (5): 0 1627--1636, 2021. doi:10.1021/acs.jcim.0c01300

    2021 doi

  5. [13]

    S. Seo, J. Lim, and W. Y. Kim. Molecular generative model via retrosynthetically prepared chemical building block assembly. Advanced Science, 10 0 (5): 0 e2206674, 2023. doi:10.1002/advs.202206674

    2023 doi

  6. [14]

    Zhavoronkov, Y

    A. Zhavoronkov, Y. A. Ivanenkov, A. Aliper, M. S. Veselov, V. A. Aladinskiy, A. V. Aladinskaya, V. A. Terentiev, S. I. Nikolenko, A. Zholus, E. Oganov, et al. Deep learning enables rapid identification of potent ddr1 kinase inhibitors. Nature Biotechnology, 37 0 (9): 0 1038--1...

    2019 doi

  7. [15]

    Zhang et al

    O. Zhang et al. Deep lead optimization: Leveraging generative ai for structural modification. Journal of the American Chemical Society, 146 0 (3): 0 31357--31370, 2024 a . doi:10.1021/jacs.3c12345

    2024 doi

  8. [16]

    Z. X. Wu et al. Leveraging language model for advanced multiproperty molecular optimization via prompt engineering. Nature Machine Intelligence, 6 0 (12): 0 1359--1369, 2024 a . doi:10.1038/s42256-024-00899-3

    2024 doi

  9. [17]

    Deep lead optimization: Leveraging generative ai for structural modification

    Odin Zhang, Haitao Lin, Hui Zhang, Huifeng Zhao, Yufei Huang, Chang-Yu Hsieh, Peichen Pan, and Tingjun Hou. Deep lead optimization: Leveraging generative ai for structural modification. Journal of the American Chemical Society, 146 0 (46): 0 31357--31370, 2024 b . doi:10.1021/...

    2024 doi

  10. [18]

    Zheng et al

    S. Zheng et al. Deep scaffold hopping with multimodal transformer neural networks. Journal of Cheminformatics, 13 0 (1): 0 87, 2021. doi:10.1186/s13321-021-00565-5

    2021 doi

  11. [19]

    C. Tan, Z. Gao, and S. Z. Li. Target-aware molecular graph generation. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Springer, 2023. doi:10.1007/978-3-031-43427-3_25

    2023 doi

  12. [20]

    T. E. Hadfield, F. Imrie, A. Merritt, K. Birchall, and C. M. Deane. Incorporating target-specific pharmacophoric information into deep generative models for fragment elaboration. Journal of Chemical Information and Modeling, 62 0 (10): 0 2280--2292, 2022. doi:10.1021/acs.jcim.1c01311

    2022 doi

  13. [21]

    Imrie, T

    F. Imrie, T. E. Hadfield, A. R. Bradley, and C. M. Deane. Deep generative design with 3d pharmacophoric constraints. Chemical Science, 12 0 (45): 0 14577--14589, 2021. doi:10.1039/D1SC02436A

    2021 doi

  14. [22]

    Skalic et al

    M. Skalic et al. From target to drug: Generative modeling for the multimodal structure-based ligand design. Molecular Pharmaceutics, 16 0 (10): 0 4282--4291, 2019. doi:10.1021/acs.molpharmaceut.9b00504

    2019 doi

  15. [23]

    X. Peng, S. Luo, J. Guan, Q. Xie, J. Peng, and J. Ma. Pocket2mol: Efficient molecular sampling based on 3d protein pockets. In Proceedings of Machine Learning Research, volume 162, pages 17644--17655, 2022

    2022

  16. [24]

    M. Wang, C. Hsieh, J. Wang, D. Wang, G. Weng, C. Shen, X. Yao, Z. Bing, H. Li, D. Cao, and T. Hou. Relation: A deep generative model for structure-based de novo drug design. Journal of Medicinal Chemistry, 65 0 (13): 0 9478--9492, 2022. doi:10.1021/acs.jmedchem.2c00732

    2022 doi

  17. [25]

    Green, D

    H. Green, D. R. Koes, and J. D. Durrant. Deepfrag: a deep convolutional neural network for fragment-based lead optimization. Chemical Science, 12 0 (23): 0 8036--8047, 2021. doi:10.1039/D1SC00163A

    2021 doi

  18. [26]

    Green and J

    H. Green and J. D. Durrant. Deepfrag: An open-source browser app for deep-learning lead optimization. Journal of Chemical Information and Modeling, 61 0 (6): 0 2523--2529, 2021. doi:10.1021/acs.jcim.1c00103

    2021 doi

  19. [27]

    Ragoza, T

    M. Ragoza, T. Masuda, and D. R. Koes. Generating 3d molecules conditional on receptor binding sites with deep generative models. Chemical Science, 13 0 (9): 0 2701--2713, 2022. doi:10.1039/D1SC06695A

    2022 doi

  20. [28]

    Masuda, M

    T. Masuda, M. Ragoza, and D. R. Koes. Generating 3d molecular structures conditional on a receptor binding site with deep generative models. arXiv preprint arXiv:2010.14442, 2020

    2010 arXiv

  21. [29]

    Liu et al

    M. Liu et al. Generating 3d molecules for target protein binding. arXiv preprint arXiv:2204.09410, 2022

    2022 arXiv

  22. [30]

    S. Luo, J. Guan, J. Ma, and J. Peng. A 3d generative model for structure-based drug design. In Advances in Neural Information Processing Systems, volume 34, pages 6229--6239, 2021

    2021

  23. [31]

    H. Lin, J. Xie, et al. Diffbp: Generative diffusion of 3d molecules for target protein binding. arXiv preprint, 2022. URL https://arxiv.org/abs/2211.11214

    2022 arXiv

  24. [32]

    Schneuing et al

    A. Schneuing et al. Structure-based drug design with equivariant diffusion models. Nature Computational Science, 4: 0 899--909, 2024

    2024

  25. [33]

    Guan et al

    J. Guan et al. 3d equivariant diffusion for target-aware molecule generation and affinity prediction. arXiv preprint, 2023. URL https://arxiv.org/abs/2303.03543

    2023 arXiv

  26. [34]

    Huang et al

    L. Huang et al. A dual diffusion model enables 3d molecule generation and lead optimization based on target pockets. Nature Communications, 15: 0 2657, 2024

    2024

  27. [35]

    Wu et al

    P. Wu et al. Guided diffusion for molecular generation with interaction prompt. Briefings in Bioinformatics, 25: 0 bbae174, 2024 b

    2024

  28. [36]

    H. Qian, W. Huang, S. Tu, and L. Xu. Kgdiff: towards explainable target-aware molecule generation with knowledge guidance. Briefings in Bioinformatics, 25: 0 bbad435, 2024

    2024

  29. [37]

    J. Xie, S. Chen, J. Lei, and Y. Yang. Diffdec: structure-aware scaffold decoration with an end-to-end diffusion model. Journal of Chemical Information and Modeling, 64: 0 2554--2564, 2024

    2024

  30. [38]

    Guan et al

    J. Guan et al. Decompdiff: diffusion models with decomposed priors for structure-based drug design. arXiv preprint, 2024. URL https://arxiv.org/abs/2403.07902

    2024 arXiv

  31. [39]

    Lu et al

    W. Lu et al. Tankbind: Trigonometry-aware neural networks for drug-protein binding structure prediction. In Advances in Neural Information Processing Systems, volume 35, pages 7236--7249, 2022

    2022

  32. [40]

    G. Landrum. RDKit: A software suite for cheminformatics, computational chemistry, and predictive modeling, volume 8. Academic Press, Cambridge, 2013

    2013

  33. [41]

    X. Peng, J. Guan, Q. Liu, and J. Ma. Moldiff: Addressing the atom-bond inconsistency problem in 3d molecule diffusion generation. arXiv preprint, 2023. URL https://arxiv.org/abs/2305.07508

    2023 arXiv

  34. [42]

    Kingma, T

    D. Kingma, T. Salimans, B. Poole, and J. Ho. Variational diffusion models. In Advances in Neural Information Processing Systems, volume 34, pages 21696--21707, 2021

    2021

  35. [43]

    Song et al

    Y. Song et al. Score-based generative modeling through stochastic differential equations. arXiv preprint, 2020. URL https://arxiv.org/abs/2011.13456

    2020 arXiv

  36. [44]

    V. G. Satorras, E. Hoogeboom, and M. Welling. E(n) equivariant graph neural networks. In Proceedings of the International Conference on Machine Learning, 2021

    2021

  37. [45]

    Guan et al

    J. Guan et al. Energy-inspired molecular conformation optimization. In International Conference on Learning Representations, 2021

    2021

  38. [46]

    Yang et al

    K. Yang et al. Analyzing learned molecular representations for property prediction. Journal of Chemical Information and Modeling, 59: 0 3370--3388, 2019

    2019

  39. [47]

    Gasteiger, J

    J. Gasteiger, J. Groß, and S. Günnemann. Directional message passing for molecular graphs. arXiv preprint, 2020. URL https://arxiv.org/abs/2003.03123

    2020 arXiv

  40. [48]

    Dhariwal and A

    P. Dhariwal and A. Nichol. Diffusion models beat gans on image synthesis. In Advances in Neural Information Processing Systems, volume 34, pages 8780--8794, 2021

    2021

  41. [49]

    D. E. Clark and S. D. Pickett. Computational methods for the prediction of ‘drug-likeness’. Drug Discovery Today, 5: 0 49--58, 2000

    2000

  42. [50]

    Ertl and A

    P. Ertl and A. Schuffenhauer. Estimation of synthetic accessibility score of drug-like molecules based on molecular complexity and fragment contributions. Journal of Cheminformatics, 1: 0 1, 2009

    2009

  43. [51]

    Sangster

    J. Sangster. Octanol-water partition coefficients of simple organic compounds. Journal of Physical and Chemical Reference Data, 18: 0 1111--1229, 1989

    1989

  44. [52]

    A. Ganesan. The impact of natural products upon modern drug discovery. Current Opinion in Chemical Biology, 12: 0 306--317, 2008

    2008

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.