Pith. sign in

REVIEW 4 major objections 4 minor 59 references

Balancing property optimization and constraint satisfaction for constrained multi-property molecular optimization

T0 review · 4 major / 4 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 A two-stage evolutionary framework balances molecular property optimization with drug-like constraints and outperforms five baselines.

desk verdict Useful extension of MOMO to constrained multi-objective optimization, but the headline performance claim is confounded by an unequal Bank-library initialization and lacks statistical grounding. read the letter →

arxiv 2411.15183 v1 pith:SOJPO55I submitted 2024-11-19 physics.chem-ph cs.AIq-bio.BM

Xin Xia , Yajie Zhang , Xiangxiang Zeng , Xingyi Zhang , Chunhou Zheng , Yansen Su This is my paper · ORCID
classification physics.chem-phcs.AIq-bio.BM
keywords constrainedmulti-objectiveoptimizationmolecularlatentspaceevolutiondrug-likeconstraintsevolutionaryalgorithmvectorfragmentationmutationrankingaggregationdrugdiscovery
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper proposes CMOMO, a constrained multi-property molecular optimization framework that treats each drug-like criterion as a constraint rather than folding it into an aggregated objective. The central claim is that by first optimizing multiple molecular properties without constraints, then gradually shifting selection pressure toward constraint satisfaction, CMOMO finds molecules that are both high-quality and feasible more often than existing methods. On two benchmark tasks and two drug-target tasks, CMOMO reports higher success rates, larger hypervolumes, and more successfully optimized molecules than five comparison methods. The practical payoff is a way to generate candidate ligands and inhibitors that respect drug-like constraints, which is what makes an optimized molecule actually usable in drug discovery.

What carries the argument

The load-bearing mechanism is the dynamic cooperative optimization loop, which alternates between a continuous implicit space of latent vectors and the discrete chemical space of SMILES strings. Offspring are generated in latent space by VFER (vector-fragmentation-based evolutionary reproduction), which applies blended linear crossover and then mutates one randomly chosen fragment of the latent vector; decoded molecules are evaluated for properties and selection feeds back into latent space. The dynamic constraint-handling strategy splits evolution into an unconstrained stage that uses non-dominated sorting and crowding distance to build convergence and diversity, and a constrained stage that ranks each candidate twice — once by properties and once by constraint dominance — then combines the two ranks with a weight α that decays from 1 to 0 across generations, so the search gradually shifts from pure property optimization to constraint satisfaction.

What would settle it

Run CMOMO's VFER strategy on a benchmark set and record, at every generation, the fraction of decoded offspring that are valid SMILES and satisfy the ring-size and toxic-substructure constraints; if offspring validity or feasibility collapses under latent-space mutation, or if CMOMO's advantage over baselines disappears when the pretrained encoder-decoder is replaced by a non-smooth representation, the central claim would be refuted.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that constrained multi-property molecular optimization is best modeled as a constrained multi-objective problem, and that a two-scenario evolutionary search can solve it better than aggregating objectives with constraints or discarding infeasible candidates. In the unconstrained scenario, CMOMO evolves latent vectors using non-dominated sorting and crowding distance to build a diverse, well-converged set; in the constrained scenario, it ranks candidates by properties and by constraint dominance and combines the ranks with a decaying weight, finally returning feasible molecules on the constrained Pareto front. Empirically, CMOMO reports the highest success rate on both benchmark tasks — 94.8% on Task 1 and 70.87% on Task 2 — and the largest mean hypervolume among the multi-objective comparison methods. On the practical tasks, it reports 75% success for β2-adrenoceptor ligands versus 59% for the best baseline, and 52.7% for GSK3β inhibitors versus at most 24.6% for baselines, with optimized molecules reaching docking energies below -7 kcal/mol. The ablation study attributes the largest performance drop to removing the vector-fragmentation-based mutation, suggesting that fragment-level mutation in the latent space is the key enabler.

Load-bearing premise

The whole search runs in a latent space produced by a pretrained encoder-decoder, and the method's success depends on that space being smooth enough that linear blends and fragment mutations decode back into valid, property-relevant molecules.

Editorial extensions

If this is right

  • If CMOMO is correct, constrained multi-property molecular optimization can be treated as a two-stage search: first build a diverse set of good-property molecules, then apply constraint pressure gradually, rather than mixing constraints into a single fitness from the start.
  • The vector-fragmentation-based mutation suggests that mutating a block of latent dimensions at once is more effective for molecule generation than single-gene mutation, reducing the search-dimension problem in long latent vectors.
  • Because the framework is agnostic to how properties are computed, its two-scenario design should transfer to other objectives, constraint sets, and property predictors without re-architecting the search.
  • The demonstrations on β2-adrenoceptor and GSK3β targets imply the method can propose candidate ligands and inhibitors that satisfy structural constraints and reach binding-energy thresholds used in drug-likeness screening.

Reading between the lines

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

  • Editorial extension: The paper's comparison does not isolate the contribution of the Bank-library initialization from the two-scenario search; a natural follow-up would vary Bank size and similarity thresholds to measure how much of the success-rate gain comes from initialization quality.
  • Editorial extension: The smooth-latent-space premise inherited from the pretrained encoder-decoder is not stress-tested in the paper; a direct experiment tracking the validity rate and property distribution of decoded offspring over generations would show how much of CMOMO's search power depends on that premise.
  • Editorial extension: The decaying weight α implements a soft transition from property preference to constraint preference, suggesting the framework could be adapted to time-varying or soft constraints by replacing the fixed schedule with a feedback controller that responds to the current feasible/infeasible ratio of the population.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The manuscript proposes CMOMO, a constrained multi-objective molecular optimization framework. CMOMO initializes a population by screening a Bank library of high-property molecules from public databases and performing linear crossover in the latent space of a pretrained encoder-decoder; it then evolves molecules using a vector fragmentation-based evolutionary reproduction (VFER) strategy. A two-stage dynamic constraint handling strategy first performs unconstrained NSGA-II-style selection and then switches to a ranking-aggregation selection that interpolates property-preference and constraint-preference ranks via a cosine-decayed weight. The framework is evaluated on four tasks (two benchmark tasks and two docking/inhibition tasks) against QMO, MolFinder, MOMO, MSO, and GB-GA-P, using success rate, number of successfully optimized molecules, hypervolume, and mean property values. The paper reports that CMOMO achieves the highest success rate and hypervolume on all four tasks.

Significance. If the comparison is fair, CMOMO is a useful contribution to constrained multi-objective molecular optimization: it combines latent-space evolutionary search with explicit constraint handling, and the two-stage ranking aggregation is a sensible design. The practical docking tasks and the case studies add applied value, and the authors promise public code. The main performance claim, however, currently rests on a comparison in which CMOMO receives an external Bank-library initialization that the baselines do not receive; until that confound is removed or controlled, the quantitative superiority claim is not established. The paper's positive aspects include the clear problem formulation, the ablation study (though only on Task 1), and the attention to practical drug-like constraints.

major comments (4)
  1. [Section 4.2 and Section 2.2.2] The claim that all methods use the same population size, number of samples, and iterations does not make the comparison fair, because CMOMO additionally receives a Bank-library initialization. In Section 4.2, CMOMO builds a Bank library of high-property molecules screened from public databases for each lead molecule and then initializes its population by linear crossover between the lead vector and each Bank vector (Fig. S6). The five baselines are not described as receiving this external high-property knowledge. If the Bank molecules already satisfy the success thresholds, the large SR and HV margins reported in Fig. 3 could be caused by starting from near-solution molecules rather than by the proposed dynamic cooperative optimization or ranking-aggregation constraint handling. The ablation in Section 2.7 removes the Bank library (CMOMO_nobank) but reports results only against CMOMO on Task 1, not against the baselines, so it does not establish that a Bank-free CMOMO still dominates. Please add a matched-initialization comparison in which baselines receive the same Bank-derived initial population, or alternatively compare CMOMO_nobank against all baselines on all tasks, and report how many Bank molecules already satisfy the success thresholds per lead.
  2. [Section 2.3 and Fig. 3] All reported performance metrics are point estimates without error bars, multiple runs, or statistical significance testing. The phrase 'significantly better performance' in Section 2.3 is therefore not supported in the statistical sense. Since the compared methods are stochastic evolutionary algorithms, run-to-run variance is expected. Please report mean and standard deviation over repeated runs, and support the SR/HV ordering with paired tests across lead molecules or bootstrapped confidence intervals.
  3. [Section 2.2.3] The success thresholds used to define SR are hand-chosen (e.g., Task 1: QED ≥ 0.85, PlogP_imp ≥ 3, Similarity ≥ 0.3; Task 2: all four scores ≥ 0.5 or 0.3), and the hypervolume reference point is set to the zero vector. Because the SR comparison depends entirely on these thresholds, the observed ordering could be an artifact of threshold placement. Please provide a sensitivity analysis over threshold values, or report a threshold-independent evaluation such as the feasible hypervolume across varying thresholds and the full property distributions of the optimized sets.
  4. [Section 4.3 and Section 4.2] The method relies on the pretrained encoder-decoder being smooth and on decoded offspring being valid and evaluable, but no validity rate is reported for CMOMO or for the baselines that also operate in the same latent space. If decoding frequently fails or produces invalid molecules, the comparison may conflate search quality with decoder behavior. Please report the fraction of generated offspring that decode to valid SMILES and the fraction that are feasible, and state how these rates compare across methods. This is important for interpreting the VFER ablation in Section 2.7, whose benefit could partly reflect differences in offspring validity rather than in search efficiency.
minor comments (4)
  1. [Introduction and Section 2.3] There are several typos: 'matrial science' in the Introduction, 'as can bee seen' in Section 2.3, and 'access the quality' where 'assess' is meant. These should be corrected in a revision.
  2. [Supplementary Materials] The text says 'A link to access the supplementary materials will be provided in the published article.' For an arXiv submission, the supplementary file should be included or the repository should contain the missing figures and experimental details (e.g., Bank library screening thresholds and the exact baselines' implementations).
  3. [Fig. 3E and 3F] The histograms are labeled 'for each lead molecule,' but the x-axis labels are difficult to parse (e.g., 'Mol 1 10050 7525'). Please provide a clearer axis label, a sorted lead index, or a caption that explains the grouping.
  4. [Section 4.4.2, Eq. (7)] The cosine decay for alpha is introduced without a sensitivity analysis. Since Eq. (7) controls the entire balance between property and constraint preferences, a brief robustness check over different schedules (e.g., fixed alpha, linear decay) would strengthen the claim that the dynamic schedule is beneficial.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: CMOMO's evaluation is benchmarked against external property predictors and independent baselines; the central performance claims are empirical, not derived from its own definitions.

full rationale

The paper's central claims are empirical comparisons on defined benchmark tasks, and the reported properties (QED, PlogP, similarity, docking scores, GSK3β inhibition) are evaluated by external predictors or widely used scoring functions, not by CMOMO's own objective definitions. The success-rate and hypervolume metrics are computed from predefined thresholds, and none of the paper's equations reduce a predicted quantity to a fitted input. The Bank-library initialization is an explicit algorithmic component, not a hidden fitted parameter; it raises a legitimate fairness/leakage concern for the comparison because the baselines do not receive this external high-property initialization, but that is an experimental-design issue rather than a circular derivation. The pretrained encoder/decoder of Winter et al. is an external dependency, and the paper's minor self-citations (e.g., to the authors' earlier MOMO work, used as a baseline and as an example of linear crossover) are not load-bearing for the main claim. No self-definitional, fitted-input-renamed-as-prediction, or uniqueness-imported-from-authors pattern is present. The derivation chain is self-contained in the sense required for circularity analysis.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim that CMOMO outperforms baselines rests on standard evolutionary machinery (NSGA-II), a pretrained latent space from prior work, a hand-designed constraint schedule, and computational property predictors. The fitted hyperparameters (alpha schedule, mutation probability, crossover parameter, fragment size, population size, number of iterations, Bank library screening) are not specified in the preprint, so the method as presented is underspecified for exact reproduction.

free parameters (6)
  • Alpha schedule in ranking aggregation = alpha = 1/2 * (1 + cos(t/T * pi))
    Hand-designed cosine decay controls the trade-off between property-preference and constraint-preference rankings; not derived from data or first principles.
  • Mutation probability pm = not specified in main text
    Controls how often a selected fragment is mutated in VFER; exact value deferred to supplementary material.
  • Blended crossover parameter d = not specified in main text
    Controls interpolation/extrapolation in linear crossover (Eq. 5); standard value not stated.
  • Fragment size in VFER = not specified in main text
    Number and length of latent vector fragments affect mutation granularity; no value provided.
  • Population size P and number of iterations T = not specified in main text
    Key evolutionary parameters are deferred to supplementary material, which is not available in the preprint.
  • Bank library size and screening thresholds = not specified in main text
    Criteria for selecting similar high-property molecules from ZINC/ChEMBL/PubChem are not given in the main text.
assumptions (5)
  • standard math NSGA-II non-dominated sorting and crowding distance correctly rank molecules in latent space
    The algorithm relies on NSGA-II's selection principles, described in Section 4.4.1, without modification.
  • domain assumption The pretrained encoder-decoder of Winter et al. creates a smooth latent space where linear interpolation and local mutation yield valid molecules
    All offspring generation and decoding depend on this latent space; invoked in Sections 4.2 and 4.3.
  • domain assumption QED, PlogP, similarity, docking score, and GSK3-beta surrogate model are adequate objective functions for drug discovery
    The benchmark and practical tasks use these computational estimators (Section 2.2.1) without independent experimental validation.
  • domain assumption The two structural constraints (ring size 5-6 and absence of 163 toxic substructures) capture key drug-likeness criteria
    Constraints are adopted from MSO (Section 2.2.1); the paper does not justify their sufficiency.
  • ad hoc to paper The alpha cosine schedule is an appropriate way to balance property and constraint preferences
    Introduced specifically for CMOMO in Eq. 7; no theoretical or empirical justification beyond ablation results.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Balancing property optimization and constraint satisfaction for constrained multi-property molecular optimization." pith.science (2026). https://pith.science/paper/SOJPO55I

@misc{pith2026241115183,
  author       = {Pith},
  title        = {Pith review of: Balancing property optimization and constraint satisfaction for constrained multi-property molecular optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SOJPO55I}},
  note         = {Machine review of arXiv:2411.15183}
}
abstract

Molecular optimization, which aims to discover improved molecules from a vast chemical search space, is a critical step in chemical development. Various artificial intelligence technologies have demonstrated high effectiveness and efficiency on molecular optimization tasks. However, few of these technologies focus on balancing property optimization with constraint satisfaction, making it difficult to obtain high-quality molecules that not only possess desirable properties but also meet various constraints. To address this issue, we propose a constrained multi-property molecular optimization framework (CMOMO), which is a flexible and efficient method to simultaneously optimize multiple molecular properties while satisfying several drug-like constraints. CMOMO improves multiple properties of molecules with constraints based on dynamic cooperative optimization, which dynamically handles the constraints across various scenarios. Besides, CMOMO evaluates multiple properties within discrete chemical spaces cooperatively with the evolution of molecules within an implicit molecular space to guide the evolutionary search. Experimental results show the superior performance of the proposed CMOMO over five state-of-the-art molecular optimization methods on two benchmark tasks of simultaneously optimizing multiple non-biological activity properties while satisfying two structural constraints. Furthermore, the practical applicability of CMOMO is verified on two practical tasks, where it identified a collection of candidate ligands of $\beta$2-adrenoceptor GPCR and candidate inhibitors of glycogen synthase kinase-3$\beta$ with high properties and under drug-like constraints.

Figures

Figures reproduced from arXiv: 2411.15183 by the authors.

Figure 1
Figure 1. Three problem models for molecular optimization and the schematic diagram of CMOMO. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. The illustrative diagram of CMOMO. (A) To begin with, CMOMO generates an initial [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Performance of CMOMO and comparison methods on two benchmark constrained multi [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Performance of CMOMO and comparison methods on Task 3. (A) Success rate (SR) [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Performance of CMOMO and comparison methods on Task 4. (A) Success rate (SR) [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Comparison of successfully optimized molecules obtained by CMOMO and three Pareto [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: Multiple evolutionary paths formed by the optimized molecules of CMOMO for a lead [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: The performance of CMOMO and its three variants on Task 1. The first variant [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: Illustrative example of the proposed fragmentation-based mutation operator. Given a [PITH_FULL_IMAGE:figures/full_fig_p019_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

59 extracted references · 57 canonical work pages

  1. [1]

    Structural optimization of drugs: Design strategies and empirical rules

    Sheng C and Li J. Structural optimization of drugs: Design strategies and empirical rules. Vol. 7. Chemical Industry Press, Beijing, 2017

    work page 2017

  2. [2]

    Applications of machine learning in drug discovery and development

    Vamathevan J, Clark D, Czodrowski P, et al. Applications of machine learning in drug discovery and development. Nature reviews Drug discovery 2019;18:463–77

    work page 2019

  3. [3]

    Beyond genera- tive models: superfast traversal, optimization, novelty, exploration and discovery (STONED) algorithm for molecules using SELFIES

    Nigam A, Pollice R, Krenn M, Passos Gomes G dos, and Aspuru-Guzik A. Beyond genera- tive models: superfast traversal, optimization, novelty, exploration and discovery (STONED) algorithm for molecules using SELFIES. Chemical science 2021;12:7079–90

    work page 2021

  4. [4]

    Computer-aided multi-objective optimization in small molecule discovery

    Fromer JC and Coley CW. Computer-aided multi-objective optimization in small molecule discovery. Patterns 2023;4

    work page 2023

  5. [5]

    Simultaneously optimizing multiple properties of β-glucosidase Bgl6 using combined (semi-) rational design strategies and investigation of the underlying mechanisms

    Li S, Cao L, Yang X, Wu X, Xu S, and Liu Y. Simultaneously optimizing multiple properties of β-glucosidase Bgl6 using combined (semi-) rational design strategies and investigation of the underlying mechanisms. Bioresource Technology 2023;374:128792

    work page 2023

  6. [6]

    Docking-based Multi-objective Molecular optimization Pipeline using Structure-constrained Genetic Algorithm

    Lee Y, Choi K, and Kim C. Docking-based Multi-objective Molecular optimization Pipeline using Structure-constrained Genetic Algorithm. In: 2022 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) . IEEE. 2022:3438–45

    work page 2022

  7. [7]

    Multi-constraint molecular generation based on conditional transformer, knowledge distillation and reinforcement learning

    Wang J, Hsieh CY, Wang M, et al. Multi-constraint molecular generation based on conditional transformer, knowledge distillation and reinforcement learning. Nature Machine Intelligence 2021;3:914–22

    work page 2021

  8. [8]

    Deep learning enables rapid identification of potent DDR1 kinase inhibitors

    Zhavoronkov A, Ivanenkov YA, Aliper A, et al. Deep learning enables rapid identification of potent DDR1 kinase inhibitors. Nature biotechnology 2019;37:1038–40

    work page 2019

Show all 59 references
  1. [9]

    Junction tree variational autoencoder for molecular graph generation

    Jin W, Barzilay R, and Jaakkola T. Junction tree variational autoencoder for molecular graph generation. In: International conference on machine learning . PMLR. 2018:2323–32

    2018

  2. [10]

    Constrained graph variational autoen- coders for molecule design

    Liu Q, Allamanis M, Brockschmidt M, and Gaunt A. Constrained graph variational autoen- coders for molecule design. Advances in neural information processing systems 2018;31

    2018

  3. [11]

    Limo: Latent inceptionism for targeted molecule generation

    Eckmann P, Sun K, Zhao B, Feng M, Gilson MK, and Yu R. Limo: Latent inceptionism for targeted molecule generation. Proceedings of machine learning research 2022;162:5777

    2022

  4. [12]

    Computational approaches streamlining drug discovery

    Sadybekov A V and Katritch V. Computational approaches streamlining drug discovery. Nature 2023;616:673–85

    2023

  5. [13]

    Artificial intelligence in drug discovery and development

    Mak KK, Wong YH, and Pichika MR. Artificial intelligence in drug discovery and development. Drug Discovery and Evaluation: Safety and Pharmacokinetic Assays 2023:1–38. 23

    2023

  6. [14]

    Chemistry42: an AI-driven platform for molec- ular design and optimization

    Ivanenkov YA, Polykovskiy D, Bezrukov D, et al. Chemistry42: an AI-driven platform for molec- ular design and optimization. Journal of Chemical Information and Modeling 2023;63:695–701

    2023

  7. [15]

    Mars: Markov molecular sampling for multi-objective drug dis- covery

    Xie Y, Shi C, Zhou H, et al. Mars: Markov molecular sampling for multi-objective drug dis- covery. In Proc. International Conference on Learning Representations 2021

    2021

  8. [16]

    Augmenting genetic algorithms with deep neural networks for exploring the chemical space

    Nigam A, Friederich P, Krenn M, and Aspuru-Guzik A. Augmenting genetic algorithms with deep neural networks for exploring the chemical space. In ICLR 2020

    2020

  9. [17]

    Guiding deep molecular optimization with genetic exploration

    Ahn S, Kim J, Lee H, and Shin J. Guiding deep molecular optimization with genetic exploration. Advances in neural information processing systems 2020;33:12008–21

    2020

  10. [18]

    Molecular de-novo design through deep reinforcement learning

    Olivecrona M, Blaschke T, Engkvist O, and Chen H. Molecular de-novo design through deep reinforcement learning. Journal of cheminformatics 2017;9:1–14

    2017

  11. [19]

    Multi-objective molecule generation using interpretable substructures

    Jin W, Barzilay R, and Jaakkola T. Multi-objective molecule generation using interpretable substructures. In: International conference on machine learning . PMLR. 2020:4849–59

    2020

  12. [20]

    Deep generative molecular design reshapes drug discovery

    Zeng X, Wang F, Luo Y, et al. Deep generative molecular design reshapes drug discovery. Cell Reports Medicine 2022

    2022

  13. [21]

    Scaffold-based molecular design with a graph generative model

    Lim J, Hwang SY, Moon S, Kim S, and Kim WY. Scaffold-based molecular design with a graph generative model. Chemical science 2020;11:1153–64

    2020

  14. [22]

    A domain-label-guided translation model for molecular optimization

    Zhang Y, Tong Y, Xia X, Wu Q, and Su Y. A domain-label-guided translation model for molecular optimization. Methods 2024

    2024

  15. [23]

    MolSearch: Search-based Multi- objective Molecular Generation and Property Optimization

    Sun M, Xing J, Meng H, Wang H, Chen B, and Zhou J. MolSearch: Search-based Multi- objective Molecular Generation and Property Optimization. In: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . 2022:4724–32

    2022

  16. [24]

    The role of ai in drug discovery: challenges, opportunities, and strategies

    Blanco-Gonzalez A, Cabezon A, Seco-Gonzalez A, et al. The role of ai in drug discovery: challenges, opportunities, and strategies. Pharmaceuticals 2023;16:891

    2023

  17. [25]

    Learning multimodal graph-to-graph translation for molecular optimization

    Jin W, Yang K, Barzilay R, and Jaakkola T. Learning multimodal graph-to-graph translation for molecular optimization. In Proc. International Conference on Learning Representations 2019

    2019

  18. [26]

    Core: Automatic molecule optimization using copy & refine strategy

    Fu T, Xiao C, and Sun J. Core: Automatic molecule optimization using copy & refine strategy. In: Proceedings of the AAAI Conference on Artificial Intelligence . Vol. 34. 01. 2020:638–45

    2020

  19. [27]

    Unpaired generative molecule-to-molecule translation for lead optimization

    Barshatski G and Radinsky K. Unpaired generative molecule-to-molecule translation for lead optimization. In: Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining . 2021:2554–64

    2021

  20. [28]

    Structure-aware conditional variational auto-encoder for constrained molecule optimization

    Yu J, Xu T, Rong Y, Huang J, and He R. Structure-aware conditional variational auto-encoder for constrained molecule optimization. Pattern Recognition 2022;126:108581

    2022

  21. [29]

    Optimizing molecules using efficient queries from property evaluations

    Hoffman SC, Chenthamarakshan V, Wadhawan K, Chen PY, and Das P. Optimizing molecules using efficient queries from property evaluations. Nature Machine Intelligence 2022;4:21–31

    2022

  22. [30]

    MolFinder: an evolutionary algorithm for the global optimization of molec- ular properties and the extensive exploration of chemical space using SMILES

    Kwon Y and Lee J. MolFinder: an evolutionary algorithm for the global optimization of molec- ular properties and the extensive exploration of chemical space using SMILES. Journal of cheminformatics 2021;13:1–14. 24

    2021

  23. [31]

    Molecule optimization via multi-objective evolutionary in implicit chemical space

    Xia X, Su Y, Zheng C, and Zeng X. Molecule optimization via multi-objective evolutionary in implicit chemical space. arXiv preprint arXiv:2212.08826 2022

    2022 arXiv

  24. [32]

    Efficient multi-objective molecular optimization in a continuous latent space

    Winter R, Montanari F, Steffen A, Briem H, No´ e F, and Clevert DA. Efficient multi-objective molecular optimization in a continuous latent space. Chemical science 2019;10:8016–24

    2019

  25. [33]

    Graph-based molecular Pareto optimisation

    Verhellen J. Graph-based molecular Pareto optimisation. Chemical Science 2022;13:7526–35

    2022

  26. [34]

    Learning continuous and data-driven molecular descriptors by translating equivalent chemical representations

    Winter R, Montanari F, No´ e F, and Clevert DA. Learning continuous and data-driven molecular descriptors by translating equivalent chemical representations. Chemical science 2019;10:1692– 701

    2019

  27. [35]

    A crossover operator using independent component analysis for real- coded genetic algorithms

    Takahashi M and Kita H. A crossover operator using independent component analysis for real- coded genetic algorithms. In: Proceedings of the 2001 congress on evolutionary computation (ieee cat. no. 01th8546) . Vol. 1. IEEE. 2001:643–9

    2001

  28. [36]

    A fast and elitist multiobjective genetic algorithm: NSGA-II

    Deb K, Pratap A, Agarwal S, and Meyarivan T. A fast and elitist multiobjective genetic algorithm: NSGA-II. IEEE transactions on evolutionary computation 2002;6:182–97

    2002

  29. [37]

    Quantifying the chemical beauty of drugs

    Bickerton GR, Paolini GV, Besnard J, Muresan S, and Hopkins AL. Quantifying the chemical beauty of drugs. Nature chemistry 2012;4:90–8

    2012

  30. [38]

    iLOGP: a simple, robust, and efficient description of n-octanol/water partition coefficient for drug design using the GB/SA approach

    Daina A, Michielin O, and Zoete V. iLOGP: a simple, robust, and efficient description of n-octanol/water partition coefficient for drug design using the GB/SA approach. Journal of chemical information and modeling 2014;54:3284–301

    2014

  31. [39]

    Why is Tanimoto index an appropriate choice for fingerprint- based similarity calculations? Journal of cheminformatics 2015;7:1–13

    Bajusz D, R´ acz A, and H´ eberger K. Why is Tanimoto index an appropriate choice for fingerprint- based similarity calculations? Journal of cheminformatics 2015;7:1–13

    2015

  32. [40]

    GuacaMol: benchmarking models for de novo molecular design

    Brown N, Fiscato M, Segler MH, and Vaucher AC. GuacaMol: benchmarking models for de novo molecular design. Journal of chemical information and modeling 2019;59:1096–108

    2019

  33. [41]

    Tartarus: A benchmarking platform for realistic and practical inverse molecular design

    Nigam A, Pollice R, Tom G, et al. Tartarus: A benchmarking platform for realistic and practical inverse molecular design. arXiv preprint arXiv:2209.12487 2022

    2022 arXiv

  34. [42]

    AutoDock Vina 1.2

    Eberhardt J, Santos-Martins D, Tillack AF, and Forli S. AutoDock Vina 1.2. 0: New dock- ing methods, expanded force field, and python bindings. Journal of chemical information and modeling 2021;61:3891–8

    2021

  35. [43]

    Multi-objective de novo drug design with conditional graph generative model

    Li Y, Zhang L, and Liu Z. Multi-objective de novo drug design with conditional graph generative model. Journal of cheminformatics 2018;10:1–24

    2018

  36. [44]

    A faster algorithm for calculating hypervolume

    While L, Hingston P, Barone L, and Huband S. A faster algorithm for calculating hypervolume. IEEE transactions on evolutionary computation 2006;10:29–38

    2006

  37. [45]

    Geo-Measures: A PyMOL plugin for protein structure ensembles analysis

    Kagami LP, Neves GM das, Timmers LFSM, Caceres RA, and Eifler-Lima VL. Geo-Measures: A PyMOL plugin for protein structure ensembles analysis. Computational Biology and Chem- istry 2020;87:107322

    2020

  38. [46]

    A knowledge-guided pre-training framework for improving molecular representation learning

    Li H, Zhang R, Min Y, Ma D, Zhao D, and Zeng J. A knowledge-guided pre-training framework for improving molecular representation learning. Nature Communications 2023;14:7568. 25

    2023

  39. [47]

    Molecular screening of glycyrrhizin- based inhibitors against ACE2 host receptor of SARS-CoV-2

    Ahmad S, Waheed Y, Abro A, Abbasi SW, and Ismail S. Molecular screening of glycyrrhizin- based inhibitors against ACE2 host receptor of SARS-CoV-2. Journal of Molecular Modeling 2021;27:206

    2021

  40. [48]

    UniProt: a worldwide hub of protein knowledge

    Consortium U. UniProt: a worldwide hub of protein knowledge. Nucleic acids research 2019;47:D506– D515

    2019

  41. [49]

    Deep evolutionary learning for molecular design

    Grantham K, Mukaidaisi M, Ooi HK, Ghaemi MS, Tchagang A, and Li Y. Deep evolutionary learning for molecular design. IEEE Computational Intelligence Magazine 2022;17:14–28

    2022

  42. [50]

    Multigranularity Surrogate Modeling for Evolutionary Multiobjective Optimization With Expensive Constraints

    Zhang Y, Jiang H, Tian Y, Ma H, and Zhang X. Multigranularity Surrogate Modeling for Evolutionary Multiobjective Optimization With Expensive Constraints. IEEE Transactions on Neural Networks and Learning Systems 2023

    2023

  43. [51]

    Design and analysis of helper-problem- assisted evolutionary algorithm for constrained multiobjective optimization

    Zhang Y, Tian Y, Jiang H, Zhang X, and Jin Y. Design and analysis of helper-problem- assisted evolutionary algorithm for constrained multiobjective optimization. Information Sci- ences 2023;648:119547

    2023

  44. [52]

    MolScore: A scoring and evaluation framework for de novo drug design

    Thomas M, O’Boyle NM, Bender A, and De Graaf C. MolScore: A scoring and evaluation framework for de novo drug design. 2024

    2024

  45. [53]

    An overview of compound prop- erties, multiparameter optimization, and computational drug design methods for PARP-1 in- hibitor drugs

    Hirlekar BU, Nuthi A, Singh KD, Murty US, and Dixit V A. An overview of compound prop- erties, multiparameter optimization, and computational drug design methods for PARP-1 in- hibitor drugs. European Journal of Medicinal Chemistry 2023;252:115300

    2023

  46. [54]

    New substructure filters for removal of pan assay interference compounds (PAINS) from screening libraries and for their exclusion in bioassays

    Baell JB and Holloway GA. New substructure filters for removal of pan assay interference compounds (PAINS) from screening libraries and for their exclusion in bioassays. Journal of medicinal chemistry 2010;53:2719–40

    2010

  47. [55]

    Privileged structures in drug discovery: medicinal chemistry and synthesis

    Yet L. Privileged structures in drug discovery: medicinal chemistry and synthesis. John Wiley & Sons, 2018

    2018

  48. [56]

    ZINC 15–ligand discovery for everyone

    Sterling T and Irwin JJ. ZINC 15–ligand discovery for everyone. Journal of chemical information and modeling 2015;55:2324–37

    2015

  49. [57]

    ChEMBL: a large-scale bioactivity database for drug discovery

    Gaulton A, Bellis LJ, Bento AP, et al. ChEMBL: a large-scale bioactivity database for drug discovery. Nucleic acids research 2012;40:D1100–D1107

    2012

  50. [58]

    PubChem substance and compound databases

    Kim S, Thiessen PA, Bolton EE, et al. PubChem substance and compound databases. Nucleic acids research 2016;44:D1202–D1213

    2016

  51. [59]

    A comprehensive review on NSGA-II for multi-objective combinatorial optimization problems

    Verma S, Pant M, and Snasel V. A comprehensive review on NSGA-II for multi-objective combinatorial optimization problems. IEEE access 2021;9:57757–91. 26

    2021

Pith tools

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