Pith. sign in

REVIEW 4 major objections 5 minor 49 references

DrugGen: Advancing Drug Discovery with Large Language Models and Reinforcement Learning Feedback

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read By fine-tuning a GPT-style molecule generator on approved drugs and then optimizing it with reinforcement learning against a predicted binding-affinity reward, the paper's DrugGen achieves 100% valid structure generation and higher…

desk verdict Useful validity improvement and clean release, but the headline affinity gain is a reward-overfitting artifact and the abstract overstates diversity/novelty. read the letter →

arxiv 2411.14157 v1 pith:WEQ6IBDJ submitted 2024-11-20 q-bio.QM cs.AI

Mahsa Sheikholeslami , Navid Mazrouei , Yousof Gheisari , Afshin Fasihi , Matin Irajpour , Ali Motahharynia This is my paper · ORCID
classification q-bio.QMcs.AI
keywords drugdiscoverylargelanguagemodelreinforcementlearningproximalpolicyoptimizationbindingaffinitypredictionSMILESgenerationmoleculardockingrepurposing
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

The paper claims that a large language model trained to write molecular formulas can be made substantially more useful for early drug discovery by first fine-tuning it on known approved drug-target pairs and then optimizing it with reinforcement learning, using a predicted binding-affinity score as the reward. The resulting model, DrugGen, produces 100% valid molecular structures versus 95.5% for its base model DrugGPT, and its molecules receive higher predicted binding affinities (median 7.22 versus 5.81) across all eight test targets. The authors argue this combination makes the generator both more reliable and more potent, and docking simulations suggest some generated molecules bind their targets more strongly than the reference drugs. If the affinity predictor's scores reflect real binding, the recipe could accelerate lead generation and drug repurposing.

What carries the argument

The load-bearing mechanism is the two-phase training loop. First, DrugGPT is supervised fine-tuned on 9,398 sequence-SMILES strings built from 1,660 approved small molecules and their targets, so the model learns the chemical language of drugs that have already passed regulatory review. Second, the fine-tuned model is run through proximal policy optimization (PPO), where each generated SMILES string receives a reward equal to PLAPT's predicted negative log affinity, multiplied by a validity check that assigns zero to invalid structures and by a factor of 0.7 if the molecule duplicates an approved drug. A KL-divergence penalty keeps the policy close to the fine-tuned model. This reward function is what simultaneously enforces chemical validity, affinity, and a balance between novelty and repurposing.

What would settle it

Take the top-scoring DrugGen molecules for ACE, NAMPT, and FABP5 and measure their actual binding (for example, IC50 or Kd) alongside an equal number of DrugGPT molecules; if DrugGen's molecules do not bind more tightly, the central affinity claim is refuted. A faster in silico check is to re-run the PPO phase with a completely different affinity predictor and see whether the apparent gain persists.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a GPT-style molecule generator can be steered toward chemically valid, high-affinity candidates by the right reward signal: supervised fine-tuning on approved drugs alone is not enough, but adding proximal policy optimization with a reward that combines a transformer-based affinity predictor (PLAPT), a rigid validity check, and a penalty for reproducing known approved molecules pushes the model to 100% validity and higher predicted affinity than the DrugGPT baseline. Across eight targets, including six with no known approved drugs, DrugGen's median PLAPT affinity was 7.22 [6.30-8.07] versus 5.81 [4.97-6.63] for DrugGPT, and for FABP5 and NAMPT the best docked poses beat the reference ligands' docking scores (FABP5/11 at -9.537 versus palmitic acid at -6.177). The paper also frames the evaluation metrics themselves—validity, diversity, novelty, and binding affinity—as a reusable standard for comparing future generative drug-design models.

Load-bearing premise

The load-bearing premise is that PLAPT's predicted binding affinity is a reliable proxy for true drug-target binding, because that same predictor is used to define the training reward and to score the final comparison; if optimizing against PLAPT merely exploits its blind spots, the reported affinity advantage would not survive contact with real assays.

Editorial extensions

If this is right

  • If the PLAPT reward transfers to real binding, DrugGen-style training should increase the fraction of generated molecules that advance to synthesis and assay, since invalid structures are almost eliminated.
  • The pipeline applies to proteins with no known approved drugs, so it could be used to generate early leads for novel or understudied targets.
  • Because the reward is pluggable, the same PPO setup could optimize other objectives—synthesis feasibility, toxicity, solubility—by swapping in the corresponding predictor.
  • Docking results like the NAMPT novel pharmacophore suggest the method can propose chemically different scaffolds that still occupy the intended binding site, which is useful for escaping crowded chemical space.
  • The evaluation metrics (validity, Tanimoto-based diversity, novelty, and binding affinity) give later models a standard way to report generation quality.

Reading between the lines

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

  • The affinity gain is measured with the same PLAPT model used as the training reward, so part of the improvement could be reward overfitting; an independent affinity predictor or wet-lab binding data would be needed to confirm the gain is real.
  • Because DrugGen's novelty is lower than DrugGPT's (41.88% versus 66.84%), fine-tuning on approved drugs seems to pull the generator toward known drug-like space; whether this is a cost depends on whether the goal is de novo scaffold discovery or repurposing.
  • Ablating the reward components (validity penalty vs affinity reward vs repetition penalty) would show which term causes the validity jump; if it is mainly the penalty, a cheaper post-hoc filter might match DrugGen without reinforcement learning.
  • The docking validation covers only four targets and a handful of molecules, so a broader docking screen over all generated molecules would reveal how often high PLAPT scores correspond to strong predicted binding geometry.
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 DrugGen, a DrugGPT-based generative model for target-specific SMILES generation. DrugGen is produced by supervised fine-tuning DrugGPT on approved drug-target pairs and then applying PPO with a reward that combines PLAPT-predicted binding affinity, an RDKit-based validity checker, and penalties for repetition and approved-drug matches. The authors evaluate DrugGen against DrugGPT on eight protein targets using validity, diversity, novelty, PLAPT affinity, and molecular docking. They report that DrugGen improves validity from 95.45% to 99.90%, produces higher predicted binding affinities (7.22 vs. 5.81), and claim that diversity and novelty are maintained; docking scores for selected DrugGen molecules are also presented. The paper includes public code and data links.

Significance. If the central claims were supported, DrugGen would be a useful, well-documented contribution to RL-based molecular generation. The validity result is a concrete, reproducible improvement: the RDKit-based validity checker is described, and the validity statistics are straightforward. The authors also share code, checkpoints, and data, which supports reproducibility. However, the headline affinity claim is not independently established: PLAPT is used both as the PPO reward and as the evaluation metric, so the higher DrugGen scores are an expected optimization outcome rather than evidence of higher true binding affinity. The docking analysis is selective and is not compared with DrugGPT. In addition, the paper's own data show statistically significant decreases in diversity and novelty, contradicting the abstract's claim that these properties are maintained. The contribution, properly scoped, may be a demonstration that PPO with PLAPT can optimize a generative model's PLAPT scores while improving validity, but the current manuscript overclaims biological and medicinal significance.

major comments (4)
  1. [Sections 4.3.3 and 4.3.4, Table 1] PLAPT's neg_log10_affinity_M is the PPO reward and also the evaluation metric for binding affinity. Since PPO explicitly maximizes this reward, the observed difference (7.22 vs. 5.81) is a check that optimization worked, not evidence that DrugGen molecules bind better in reality. To support the affinity claim, the authors need an independent evaluation: a held-out or different affinity predictor, docking of both DrugGen and DrugGPT outputs under identical protocols, or experimental data. The Discussion's caveat that the reward model has inherent accuracy and specificity limitations does not address this circularity.
  2. [Abstract, Section 2.2, Figures 3A and 3B] The abstract states that DrugGen maintains diversity and novelty, but the reported values show large significant declines: diversity falls from 84.54% to 60.32% and novelty from 66.84% to 41.88%, both with highly significant test statistics. The text acknowledges these decreases, calling generated molecules more similar and fewer novel, yet concludes there is a good balance. No criterion or comparison is provided for what counts as maintained or balanced, so the claim is unsupported and the abstract is misleading.
  3. [Section 2.3, Table 2, Methods 4.3.4] The docking evidence is not a controlled comparison. Only DrugGen molecules are docked; DrugGPT molecules are not docked with the same protocol, and only a small subset of the 122 docked molecules is reported in Table 2 without a stated selection criterion. The ACE results are acknowledged to bind different sites from the reference, so they cannot be read as target-site validation, and the FABP5 and NAMPT examples are anecdotal. Moreover, the method text calls the docking blind, while Table 3 lists grid boxes of 30-40 Å, which is not a whole-protein search; the exact search space needs to be clarified. Thus the docking section cannot substitute for an independent affinity assessment.
  4. [Section 2.3, Table 1] The aggregate PLAPT comparison pools eight targets, of which one (FABP5) shows no significant difference after Bonferroni correction, and the Discussion concedes variability across targets. Reporting a pooled median across targets hides per-target differences and overstates consistency. The authors should present per-target effect sizes and interpret the aggregate statistic with the non-significant target explicitly excluded or modeled.
minor comments (5)
  1. [Throughout] There are several textual typos, including 'DrugGen/quotesingle.ts1' in Sections 2.3 and 3 and inconsistent spacing in 'F ABP5'; these should be corrected.
  2. [Abstract and Section 2.2] The validity number is reported inconsistently: the abstract and novelty section say 100% valid, while Section 2.2 reports 99.90% for DrugGen; the manuscript should state which generation set each number refers to.
  3. [Section 2.2 and Table 1] The p-values are reported as 'P = 0' for diversity and as 'P = 1' for FABP5; exact p-values should be given as inequalities (for example, P < 10^-300) rather than 0, and the FABP5 value should be reported as non-significant after correction.
  4. [Section 4.3.3 and References] PLAPT is cited only as a 2024 bioRxiv preprint; since it is central to both the method and the evaluation, the authors should cite the version used and specify whether it was used with default weights or fine-tuned.
  5. [Figure 1] The caption of Figure 1 says the assessment includes binding affinity for both DrugGen and DrugGPT, but docking simulations are only reported for DrugGen; the figure and text should be aligned.

Circularity Check

2 steps flagged · score 6.0 of 10

Headline affinity advantage is an evaluation of the PPO reward itself: PLAPT is both the training reward and the reported affinity metric, so the 7.22 vs 5.81 comparison largely measures reward optimization; the docking check is partial and does not compare DrugGPT.

  1. fitted input called prediction [Section 4.3.3 (Reward Function) and Section 4.3.4 (PLAPT Binding Affinity Assessment); also Abstract and Section 2.3]
    "The affinities of the generated structures with their respective targets were evaluated using PLAPT’s neg_log10_affinity_M output. ... The same set of molecules generated during the novelty assessment was used to evaluate the binding affinities of the compounds produced by DrugGPT and DrugGen."

    PPO is trained to maximize exactly this PLAPT neg_log10_affinity_M reward (with validity and repetition penalties). The headline metric, 'predicted binding affinities' (7.22 vs 5.81), is then measured with the same PLAPT model. Therefore the DrugGen advantage is a measure of how well PPO optimized its reward, not an independent test of real binding affinity. The evaluation oracle is identical to the training oracle by construction, so the result cannot falsify the reward design or show that DrugGen makes genuinely better ligands. The docking simulations are a partial external check, but they are reported only for DrugGen and include no parallel DrugGPT docking scores or statistical comparison.

  2. self citation load bearing [Section 2.2 and Section 4.3.4 (DrugGen Assessment, target selection)]
    "six proteins without known approved drugs, i.e., GALM, FB5L3, OPSB, NAMPT, PGK2, and FABP5, that identified as having a high probability of being targeted by approved small molecules through our newly developed druggability scoring algorithm, DrugTar [24]. ... The selection of these six targets was according to our recent study 'DrugTar Improves Druggability Prediction by Integrating Large Language Models and Gene Ontologies' [24]."

    The claim that these six proteins are druggable, previously untargeted proteins rests on the authors' own bioRxiv preprint DrugTar, whose authors overlap with the present paper. This self-citation is load-bearing for the paper's framing that DrugGen addresses 'previously untargeted or unexplored therapeutic areas.' If DrugTar's prioritization is not independently valid, the target selection loses its stated rationale. This does not determine the affinity numbers, but it is a load-bearing self-citation for the novelty and unexplored-target claims.

full rationale

The central circularity is that PLAPT supplies both the PPO training reward (Section 4.3.3) and the evaluation metric for binding affinity (Section 4.3.4). Reporting higher PLAPT scores for DrugGen is an expected consequence of optimizing against PLAPT, not independent evidence of improved binding; the abstract's phrase 'higher predicted binding affinities' is honest about being predictions, but the predictions are made by the very model whose score was maximized. Docking provides some external support, yet it is limited to DrugGen, covers only four targets, and lacks a DrugGPT docking comparison or statistical tests, so it does not break the circularity of the headline affinity claim. The paper's own diversity and novelty numbers also contradict the abstract's 'maintaining diversity and novelty': DrugGen diversity drops from 84.54% to 60.32% and novelty from 66.84% to 41.88%, both highly significant—this is an internal inconsistency rather than a circular derivation, but it further undermines the balanced-claims framing. The DrugTar self-citation for target selection is a separate load-bearing self-reference, though it is not the source of the affinity numbers. Overall, the main quantified claim is partially circular by construction, while the docking validation and the deterministic validity checker keep the paper from being entirely self-referential; hence a score of 6 rather than 8 or 10.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the domain assumption that PLAPT scores reflect real binding and that approved drugs are a good training prior. The only explicitly stated numeric free parameter is the 0.7 penalty for regenerating approved drugs; other PPO hyperparameters are either standard or unreported. No new entities are introduced.

free parameters (2)
  • Reward penalty factor for approved-drug matches = 0.7
    Hand-chosen multiplicative penalty applied to the PPO reward when a generated SMILES matches a molecule in the approved-drug set (Section 4.3.3).
  • PPO KL divergence coefficient = Not reported
    Controls how strongly the optimized policy is anchored to the reference model; its value affects the validity-reward trade-off but is not stated in the methods.
assumptions (4)
  • domain assumption Approved drugs constitute a useful training distribution for generating new drug-like molecules
    The model is fine-tuned on approved drug-target pairs, assuming this imparts drug-like properties to generated molecules (Section 4.1, Discussion).
  • domain assumption PLAPT binding-affinity scores correlate with true ligand-target binding
    PLAPT is used both as the reward signal (Section 4.3.3) and as the evaluation metric (Section 4.3.4); if this correlation is weak, the affinity improvement is an artifact.
  • domain assumption GLIDE XP docking scores are reliable for ranking generated molecules
    Docking is used to validate generated molecules (Section 4.3.4); reliability is asserted from redocking RMSD values but no benchmark against known actives is provided.
  • standard math Standard machine learning and statistical methods apply (PPO, Tanimoto similarity, Mann-Whitney U)
    The paper relies on established algorithms and tests without derivation (Sections 4.3.2 and 4.3.4).

how reviews work

0 comments
Cite this review

Pith. "Pith review of DrugGen: Advancing Drug Discovery with Large Language Models and Reinforcement Learning Feedback." pith.science (2026). https://pith.science/paper/WEQ6IBDJ

@misc{pith2026241114157,
  author       = {Pith},
  title        = {Pith review of: DrugGen: Advancing Drug Discovery with Large Language Models and Reinforcement Learning Feedback},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WEQ6IBDJ}},
  note         = {Machine review of arXiv:2411.14157}
}
read the original abstract

Traditional drug design faces significant challenges due to inherent chemical and biological complexities, often resulting in high failure rates in clinical trials. Deep learning advancements, particularly generative models, offer potential solutions to these challenges. One promising algorithm is DrugGPT, a transformer-based model, that generates small molecules for input protein sequences. Although promising, it generates both chemically valid and invalid structures and does not incorporate the features of approved drugs, resulting in time-consuming and inefficient drug discovery. To address these issues, we introduce DrugGen, an enhanced model based on the DrugGPT structure. DrugGen is fine-tuned on approved drug-target interactions and optimized with proximal policy optimization. By giving reward feedback from protein-ligand binding affinity prediction using pre-trained transformers (PLAPT) and a customized invalid structure assessor, DrugGen significantly improves performance. Evaluation across multiple targets demonstrated that DrugGen achieves 100% valid structure generation compared to 95.5% with DrugGPT and produced molecules with higher predicted binding affinities (7.22 [6.30-8.07]) compared to DrugGPT (5.81 [4.97-6.63]) while maintaining diversity and novelty. Docking simulations further validate its ability to generate molecules targeting binding sites effectively. For example, in the case of fatty acid-binding protein 5 (FABP5), DrugGen generated molecules with superior docking scores (FABP5/11, -9.537 and FABP5/5, -8.399) compared to the reference molecule (Palmitic acid, -6.177). Beyond lead compound generation, DrugGen also shows potential for drug repositioning and creating novel pharmacophores for existing targets. By producing high-quality small molecules, DrugGen provides a high-performance medium for advancing pharmaceutical research and drug discovery.

Figures

Figures reproduced from arXiv: 2411.14157 by the authors.

Figure 1
Figure 1. Schematic representation of model development and e [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Training process of DrugGen. (A) Learning curve of th [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Comparison of molecular diversity, novelty, and bin [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Visualization of ligand binding in active sites acro [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Comparison of pharmacophores of NAMPT inhibitors in [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 36 canonical work pages

  1. [1]

    AI-enabled organoids: construction, analysi s, and application,

    Bai, Long and Wu, Yan and Li, Guangfeng and Zhang, Wencai a nd Zhang, Hao and Su, Jiacan, "AI-enabled organoids: construction, analysi s, and application," Bioactive Materials, vol. 31, pp. 525–548, 2024

    work page 2024

  2. [2]

    Defining and Exploring Chemical Space s,

    Coley, Connor W., "Defining and Exploring Chemical Space s," Trends in Chemistry , vol. 3, no. 2, pp. 133–145, 2021. DOI: 10.1016/j.trechm.202 0.11.004

    work page doi:10.1016/j.trechm.202 2021

  3. [3]

    Why 90% of clinical drug development fails and how to improve it?,

    Sun, Duxin and Gao, Wei and Hu, Hongxiang and Zhou, Simon, "Why 90% of clinical drug development fails and how to improve it?," Acta Pharmaceutica Sinica B , vol. 12, no. 7, pp. 3049–3062, 2022. DOI: 10.1016/j.apsb.2022.02.0 02

    work page doi:10.1016/j.apsb.2022.02.0 2022

  4. [4]

    Generative Models for de Novo Drug Design,

    Tong, Xiaochu and Liu, Xiaohong and Tan, Xiaoqin and Li, X utong and Jiang, Ji- axin and Xiong, Zhaoping and Xu, Tingyang and Jiang, Hualian g and Qiao, Nan and Zheng, Mingyue, "Generative Models for de Novo Drug Design, " Journal of Medicinal Chemistry, vol. 64, no. 19, pp. 14011–14027, 2021. DOI: 10.1021/acs.j medchem.1c00927

    doi:10.1021/acs.j 2021

  5. [5]

    Deep generative molecular design res hapes drug discovery,

    Zeng, Xiangxiang and Wang, Fei and Luo, Yuan and gu Kang, S eung and Tang, Jian and Lightstone, Felice C. and Fang, Evandro F. and Cornell, W endy and Nussinov, Ruth and Cheng, Feixiong, "Deep generative molecular design res hapes drug discovery," Cell Reports Medicine, vol. 3, no. 12, 2022. DOI: 10.1016/j.xcrm.2022.100794

    arXiv 2022

  6. [6]

    D e novo molecular design and generative models,

    Meyers, Joshua and Fabian, Benedek and Brown, Nathan, "D e novo molecular design and generative models," Drug Discovery Today , vol. 26, no. 11, pp. 2707–2715, 2021. DOI: https://doi.org/10.1016/j.drudis.2021.05.019

    work page doi:10.1016/j.drudis.2021.05.019 2021

  7. [7]

    De novo generation of hit-like molecul es from gene expression signatures using artificial intelligence,

    Méndez-Lucio, Oscar and Baillif, Benoit and Clevert, Dj ork Arné and Rouquié, David and Wichard, Joerg, "De novo generation of hit-like molecul es from gene expression signatures using artificial intelligence," Nature Communications, vol. 11, no. 1, pp. 10,

    work page

  8. [8]

    A Generative and Causal Pharmacokine tic Model for Factor VIII in Hemophilia A: A Machine Learning Framework for Conti nuous Model Refine- ment,

    Janssen, Alexander and Smalbil, Louk and Bennis, Frank C . and Cnossen, Marjon H. and Mathôt, Ron A.A., "A Generative and Causal Pharmacokine tic Model for Factor VIII in Hemophilia A: A Machine Learning Framework for Conti nuous Model Refine- ment," Clinical Pharmacology and Therapeutics, vol. 115, no. 4, pp. 881–889, 2024. DOI: 10.1002/cpt.3203

    work page doi:10.1002/cpt.3203 2024

Show all 49 references
  1. [9]

    Application of machine learning techniques to the analysis and prediction of drug pharmacokinetics,

    Ota, Ryosaku and Yamashita, Fumiyoshi, "Application of machine learning techniques to the analysis and prediction of drug pharmacokinetics," Journal of Controlled Release , vol. 352, pp. 961–969, 2022. DOI: 10.1016/j.jconrel.2022. 11.014

    2022 doi

  2. [10]

    Using Generative Mod eling to Endow with Po- tency Initially Inert Compounds with Good Bioavailability and Low Toxicity,

    Horne, Robert I. and Wilson-Godber, Jared and González Díaz, Alicia and Brotza- kis, Z. Faidon and Seal, Srijit and Gregory, Rebecca C. and Po ssenti, Andrea and Chia, Sean and Vendruscolo, Michele, "Using Generative Mod eling to Endow with Po- tency Initially Inert Compounds ...

    2024 doi

  3. [11]

    Revolutionizin g pharmacokinetics: the dawn of AI-powered analysis,

    Ghayoor, Ali and Kohan, Hamed Gilzad, "Revolutionizin g pharmacokinetics: the dawn of AI-powered analysis," Journal of Pharmacy & Pharmaceutical Sciences , vol. 27, pp. 12671, 2024

    2024

  4. [12]

    Using Domain-Specifi c Fingerprints Generated through Neural Networks to Enhance Ligand-Based Virtual Sc reening,

    Menke, Janosch and Koch, Oliver, "Using Domain-Specifi c Fingerprints Generated through Neural Networks to Enhance Ligand-Based Virtual Sc reening," Journal of Chemical Information and Modeling , vol. 61, no. 2, pp. 664–675, 2021. DOI: 10.1021/acs.jcim.0c01208

    2021 doi

  5. [13]

    AI in drug discovery and its c linical relevance,

    Qureshi, Rizwan and Irfan, Muhammad and Gondal, Taimoo r Muzaffar and Khan, She- heryar and Wu, Jia and Hadi, Muhammad Usman and Heymach, John and Le, Xiuning and Yan, Hong and Alam, Tanvir, "AI in drug discovery and its c linical relevance," Heliyon, vol. 9, no. 7, 2023. DOI: ...

    2023 doi

  6. [14]

    Deep learning for dr ug discovery: A study of identi- fying high efficacy drug compounds using a cascade transfer le arning approach,

    Zhuang, Dylan and Ibrahim, Ali K., "Deep learning for dr ug discovery: A study of identi- fying high efficacy drug compounds using a cascade transfer le arning approach," Applied Sciences (Switzerland) , vol. 11, no. 17, pp. 7772, 2021. DOI: 10.3390/app11177772

    2021 doi

  7. [15]

    Unleashing the power of generative AI in drug discovery,

    Gangwal, Amit and Lavecchia, Antonio, "Unleashing the power of generative AI in drug discovery," Drug Discovery Today , vol. 29, no. 6, pp. 103992, 2024. DOI: 10.1016/j.drudis.2024.103992

    2024

  8. [16]

    PLAPT: Protein-Ligand Binding Affinity Prediction Using Pretraine d Transformers,

    Rose, Tyler and Monti, Nicolò and Anand, Navvye and Shen , Tianyu, "PLAPT: Protein-Ligand Binding Affinity Prediction Using Pretraine d Transformers," bioRxiv, pp. 2024.02.08.575577, 2024

    2024

  9. [17]

    Pro- Gen: Language Modeling for Protein Generation,

    Madani, Ali and McCann, Bryan and Naik, Nikhil and Keska r, Nitish Shirish and Anand, Namrata and Eguchi, Raphael R. and Huang, Po-Ssu and S ocher, Richard, "Pro- Gen: Language Modeling for Protein Generation," arXiv preprint arXiv:2004.03497 , 2020

    2004 arXiv

  10. [18]

    Language models are few -shot learners,

    Brown, Tom B. and Mann, Benjamin and Ryder, Nick and Subb iah, Melanie and Ka- plan, Jared and Dhariwal, Prafulla and Neelakantan, Arvind and Shyam, Pranav and Sastry, Girish and Askell, Amanda and Agarwal, Sandhini and Herbert-Voss, Ariel and Krueger, Gretchen and Henighan, T...

    2020

  11. [19]

    DrugGPT: A GPT-based Strategy for Desi gning Potential Ligands Targeting Specific Proteins,

    Li, Yuesen and Gao, Chengyi and Song, Xin and Wang, Xiang yu and Xu, Yun- gang and Han, Suxia, "DrugGPT: A GPT-based Strategy for Desi gning Potential Ligands Targeting Specific Proteins," bioRxiv, pp. 2023.06.29.543848, 2023. DOI: 10.1101/2023.06.29.543848

    2023 doi

  12. [20]

    SMILES, a Chemical Language and Inf ormation System: 1: Intro- duction to Methodology and Encoding Rules,

    Weininger, David, "SMILES, a Chemical Language and Inf ormation System: 1: Intro- duction to Methodology and Encoding Rules," Journal of Chemical Information and Computer Sciences , vol. 28, no. 1, pp. 31–36, 1988. DOI: 10.1021/ci00057a005. 17

    1988 doi

  13. [21]

    Auto regressive models for gene reg- ulatory network inference: Sparsity, stability and causal ity issues,

    Michailidis, George and D’Alché-Buc, Florence, "Auto regressive models for gene reg- ulatory network inference: Sparsity, stability and causal ity issues," Mathematical Bio- sciences, vol. 246, no. 2, pp. 326–334, 2013. DOI: 10.1016/j.mbs.201 3.10.003

    2013 doi

  14. [22]

    Refining dataset curation methods for deep learning-based automated tuberculosis screening,

    Kim, Tae Kyung and Yi, Paul H. and Hager, Gregory D. and Li n, Cheng Ting, "Refining dataset curation methods for deep learning-based automated tuberculosis screening," Journal of Thoracic Disease , vol. 12, no. 9, pp. 5078–5085, 2020. DOI: 10.21037/jtd.2019.08.34

    2020 doi

  15. [23]

    A Deep Learning App roach to Antibiotic Dis- covery,

    Stokes, Jonathan M. and Yang, Kevin and Swanson, Kyle an d Jin, Wengong and Cubillos-Ruiz, Andres and Donghia, Nina M. and MacNair, Cra ig R. and French, Shawn and Carfrae, Lindsey A. and Bloom-Ackerman, Zohar and Tran, Victoria M. and Chiappino-Pepe, Anush and Badran, Ahmed H...

    2020

  16. [24]

    DrugTar Improves Druggability Pred iction by Integrating Large Language Models and Gene Ontologies,

    Borhani, Niloofar and Izadi, Iman and Motahharynia, Al i and Sheikholeslami, Mahsa and Gheisari, Yousof, "DrugTar Improves Druggability Pred iction by Integrating Large Language Models and Gene Ontologies," bioRxiv, pp. 2024.09.21.614218, 2024. DOI: 10.1101/2024.09.21.614218

    2024 doi

  17. [25]

    Properties of FDA-approved small m olecule protein kinase in- hibitors: A 2024 update,

    Roskoski, Robert, "Properties of FDA-approved small m olecule protein kinase in- hibitors: A 2024 update," Pharmacological Research, vol. 200, pp. 107059, 2024. DOI: 10.1016/j.phrs.2024.107059

    2024

  18. [26]

    Physicochemical Properties a nd Pharmacokinetics,

    Loftsson, Thorsteinn, "Physicochemical Properties a nd Pharmacokinetics," pp. 85–104,

  19. [27]

    Chapter 1 - Introduction,

    Di, Li and Kerns, Edward H., "Chapter 1 - Introduction," in Drug-Like Properties (Sec- ond Edition) , 2nd ed., L. Di and E. H. Kerns, Eds. Boston: Academic Press, 2 016, pp. 1–3. DOI: https://doi.org/10.1016/B978-0-12-801076-1.00001-0 . A vailable at: https://www.sciencedirect...

    doi

  20. [28]

    DrugMetric: quantitative drug-likeness scoring based on chemical space distance,

    Li, Bowen and Wang, Zhen and Liu, Ziqi and Tao, Yanxin and Sha, Chulin and He, Min and Li, Xiaolin, "DrugMetric: quantitative drug-likeness scoring based on chemical space distance," Briefings in Bioinformatics , vol. 25, no. 4, 2024. DOI: 10.1093/bib/bbae321

    2024 doi

  21. [29]

    Drug Repurposing: An Effective Tool in Modern D rug Discovery,

    Kulkarni, V. S. and Alagarsamy, V. and Solomon, V. R. and Jose, P. A. and Mu- rugesan, S., "Drug Repurposing: An Effective Tool in Modern D rug Discovery," Russian Journal of Bioorganic Chemistry , vol. 49, no. 2, pp. 157–166, 2023. DOI: 10.1134/S1068162023020139

    2023 doi

  22. [30]

    Angiotensin-converting enzyme open for business: s tructural insights into the subdomain dynamics,

    Cozier, Gyles E. and Lubbe, Lizelle and Sturrock, Edwar d D. and Acharya, K. Ravi, "Angiotensin-converting enzyme open for business: s tructural insights into the subdomain dynamics," FEBS Journal , vol. 288, no. 7, pp. 2238–2256, 2021. DOI: 10.1111/febs.15601

    2021 doi

  23. [31]

    DrugBank: a comprehensive resource for in silico drug disc overy and exploration.,

    Wishart, David S. and Knox, Craig and Guo, An Chi and Shri vastava, Savita and Hassanali, Murtaza and Stothard, Paul and Chang, Zhan and Wo olsey, Jennifer, 18 "DrugBank: a comprehensive resource for in silico drug disc overy and exploration.," Nucleic acids research , vol. 34...

    2006 doi

  24. [32]

    UniProt: the Universal Prote in Knowledgebase in 2023. Nucleic Acids Res. 51:D523–D531 (2023),

    The UniProt Consortium, "UniProt: the Universal Prote in Knowledgebase in 2023. Nucleic Acids Res. 51:D523–D531 (2023)," Nucleic acids research, vol. 51, no. November 2022, pp. 523–531, 2023

    2023

  25. [33]

    Basic local alignment search tool,

    Altschul, Stephen F. and Gish, Warren and Miller, Webb a nd Myers, Eugene W. and Lipman, David J., "Basic local alignment search tool," Journal of Molecular Biology , vol. 215, no. 3, pp. 403–410, 1990. DOI: 10.1016/S0022-2836 (05)80360-2

    1990 doi

  26. [34]

    The ChE MBL Database in 2023: A drug discovery platform spanning multiple bioact ivity data types and time periods,

    Zdrazil, Barbara and Felix, Eloy and Hunter, Fiona and M anners, Emma J and Black- shaw, James and Corbett, Sybilla and de Veij, Marleen and Ioa nnidis, Harris and Mendez Lopez, David and Mosquera, Juan F and Others, "The ChE MBL Database in 2023: A drug discovery platform spa...

    2023

  27. [35]

    ZINC20 - A Free Ultralarge-Scale C hemical Database for Ligand Discovery,

    Irwin, John J. and Tang, Khanh G. and Young, Jennifer and Dandarchuluun, Chinzorig and Wong, Benjamin R. and Khurelbaatar, Munkhzul and Moroz, Yurii S. and Mayfield, John and Sayle, Roger A., "ZINC20 - A Free Ultralarge-Scale C hemical Database for Ligand Discovery," Journal of...

    2020 doi

  28. [36]

    A vailable at: https://huggingface.co/docs/trl/en/index

    Transformer Reinforcement Learning , [cited 2024]. A vailable at: https://huggingface.co/docs/trl/en/index

    2024

  29. [37]

    A vailable at:https://github.com/openai/summarize-from-feedback

    Summarize from Feedback. A vailable at:https://github.com/openai/summarize-from-feedback

  30. [38]

    A vailable at: https://www.rdkit.org/

    RDKit, [cited 2024]. A vailable at: https://www.rdkit.org/

    2024

  31. [39]

    DisGeNET: A comprehensive pl atform integrating infor- mation on human disease-associated genes and variants,

    Piñero, Janet and Bravo, Álex and Queralt-Rosinach, Nú ria and Gutiérrez-Sacristán, Alba and Deu-Pons, Jordi and Centeno, Emilio and García-Gar cía, Javier and Sanz, Ferran and Furlong, Laura I., "DisGeNET: A comprehensive pl atform integrating infor- mation on human disease-a...

    2017 doi

  32. [40]

    Wh y is Tanimoto index an appropriate choice for fingerprint-based similarity calcu lations?,

    Bajusz, Dávid and Rácz, Anita and Héberger, Károly, "Wh y is Tanimoto index an appropriate choice for fingerprint-based similarity calcu lations?," Journal of Chemin- formatics, vol. 7, no. 1, pp. 1–13, 2015. DOI: 10.1186/s13321-015-006 9-3

    2015 doi

  33. [41]

    The Generation of a Unique Machine Descr iption for Chemical Struc- tures—A Technique Developed at Chemical Abstracts Service ,

    Morgan, H. L., "The Generation of a Unique Machine Descr iption for Chemical Struc- tures—A Technique Developed at Chemical Abstracts Service ," Journal of Chemical Documentation, vol. 5, no. 2, pp. 107–113, 1965. DOI: 10.1021/c160017a018

    1965 doi

  34. [42]

    Protein-Ligand Blind Docking Using QuickVina-W with Inter-Process Spatio-Temporal Integration,

    Hassan, Nafisa M. and Alhossary, Amr A. and Mu, Yuguang an d Kwoh, Chee Keong, "Protein-Ligand Blind Docking Using QuickVina-W with Inter-Process Spatio-Temporal Integration," Scientific Reports , vol. 7, no. 1, pp. 15451, 2017. DOI: 10.1038/s41598-017- 15571-7. 19

    2017 doi

  35. [43]

    Protein and ligand preparation: Parameters, protocols, and influence on virtual screening enrichments,

    Madhavi Sastry, G. and Adzhigirey, Matvey and Day, Tyle r and Annabhimoju, Ramakr- ishna and Sherman, Woody, "Protein and ligand preparation: Parameters, protocols, and influence on virtual screening enrichments," Journal of Computer-Aided Molecular Design, vol. 27, no. 3, pp. ...

    2013 doi

  36. [44]

    Schrödinger Release 2024–2: LigPrep , 2024, Schrödinger, LLC: New York, NY

    2024

  37. [45]

    OPLS4: Improving force field accuracy on challenging regimes of chemical space,

    Lu, Chao and Wu, Chuanjie and Ghoreishi, Delaram and Che n, Wei and Wang, Lingle and Damm, Wolfgang and Ross, Gregory A. and Dahlgren, Markus K. and Russell, Ellery and Von Bargen, Christopher D. and Abel, Robert and Fr iesner, Richard A. and Harder, Edward D., "OPLS4: Improvi...

    2021 doi

  38. [46]

    and Banks, Jay L

    Friesner, Richard A. and Banks, Jay L. and Murphy, Rober t B. and Halgren, Thomas A. and Klicic, Jasna J. and Mainz, Daniel T. and Repasky, Matt hew P. and Knoll, Eric H. and Shelley, Mee and Perry, Jason K. and Shaw, David E. and F rancis, Perry and Shenkin, Peter S., "Glide:...

  39. [49]

    Method and Assessment of Docking Accuracy," Journal of Medicinal Chemistry , vol. 47, no. 7, pp. 1739–1749, 2004. DOI: 10.1021/jm0306430. 20

    2004 doi

  40. [2015]

    DOI: 10.1016/b978-0-12-801411-0.00003-2

    doi

  41. [2020]

    DOI: 10.1038/s41467-019-13807-w

    doi

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.