pith. sign in

arxiv: 2606.26657 · v1 · pith:BURCVHA4new · submitted 2026-06-25 · 💻 cs.LG

Target-Aware Bandit Allocation for Scalable Surrogate Optimization in Chemical Space

Mohammad Haddadnia , Yuvan Chali , Abhilash Jayaraj , Constance Kraay , Joana Reis , Felix Strieth-Kalthoff , Haribabu Arthanari This is my paper

Pith reviewed 2026-06-26 05:26 UTC · model grok-4.3

classification 💻 cs.LG
keywords bandit optimizationsurrogate-based optimizationchemical spacedrug discoverymulti-armed banditsscalable screeningvirtual screening
0
0 comments X

The pith

Bandit allocation of surrogate inference across chemical partitions enables optimization over libraries too large for full evaluation.

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

The paper introduces BOBa as a way to avoid running surrogate models on every compound in libraries of billions or trillions of molecules. It partitions the chemical action space and treats each partition as an arm in a multi-armed bandit, then uses optimism-under-uncertainty policies to decide where to spend inference and evaluation effort. The approach concentrates resources on regions that show early promise while still exploring. Real-library experiments indicate that this bandit-driven allocation produces a workable balance between finding high-utility candidates and keeping computational cost manageable. The central result is that meaningful partitions plus bandit exploration are required for the method to outperform simpler allocation schemes.

Core claim

By modeling partitions of the chemical action space as arms in a multi-armed bandit and applying optimism-under-uncertainty selection, BOBa adaptively allocates surrogate inference and downstream evaluations only to empirically promising partitions, removing the need to perform full-library inference while still identifying high-utility candidates from synthesis-on-demand libraries.

What carries the argument

The BOBa framework, which partitions the action space and treats each partition as a bandit arm whose reward signal guides allocation of surrogate computations.

If this is right

  • Surrogate inference cost drops while screening performance remains competitive with full-library methods.
  • Virtual screening becomes practical for current synthesis-on-demand libraries of billions to trillions of compounds.
  • A tunable tradeoff appears between the fraction of partitions explored and the quality of the final candidate set.
  • Optimism-under-uncertainty bandit policies are required for effective concentration of computation on high-utility regions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same partition-and-bandit pattern could be tested on other large discrete spaces such as protein design or materials libraries.
  • The quality of the initial partitioning step directly limits how quickly the bandit can locate good regions.
  • Combining the allocation layer with more expressive surrogates or active-learning loops could further reduce the number of expensive evaluations needed.

Load-bearing premise

The chemical action space admits a partitioning into arms such that partial observations from the bandit process can reliably identify and concentrate on high-utility regions without systematic bias from the partition boundaries.

What would settle it

A controlled experiment on a library whose highest-utility compounds lie entirely inside one partition, yet the bandit consistently under-allocates inference to that partition after initial observations, producing lower final performance than uniform allocation across all partitions.

Figures

Figures reproduced from arXiv: 2606.26657 by Abhilash Jayaraj, Constance Kraay, Felix Strieth-Kalthoff, Haribabu Arthanari, Joana Reis, Mohammad Haddadnia, Yuvan Chali.

Figure 1
Figure 1. Figure 1: Evaluation of Bandit algorithms, and observed tradeoff between performance and inference cost. (Left) Optimization trajectories of BOBA with different Bandit algorithms as a function of the number of costly black-box evaluations. BO with full-library inference (maroon) is included as an upper boundary. Performance is quantified by the number of retrieved molecules from the top-100 candidates from the full … view at source ↗
Figure 2
Figure 2. Figure 2: Influence of static vs. dynamic partitioning. Opti￾mization trajectories of BOBA with UCB1 using K-Means-based partitions, compared to setting in which a partition for inference is randomly selected at each iteration. All experiments are reported on the Enamine-5M library docked against CKB. Trajectories are shown as the mean over 5 independent runs from different seed populations. The shaded area indicate… view at source ↗
Figure 4
Figure 4. Figure 4: Effect of feature space on end-to-end BOBA per￾formance. Optimization trajectories of BOBA, with partitions constructed using T5Chem language model embeddings, or us￾ing unrefined physicochemical descriptors. All experiments are reported on the Enamine-5M library docked against CKB. Tra￾jectories are shown as the mean over 5 independent runs from different seed populations. The shaded area indicates the st… view at source ↗
Figure 5
Figure 5. Figure 5: BOBA across targets and difficulty regimes. Opti￾mization trajectories of BOBA on different optimization problems. Trajectories are shown as the mean over 5 independent runs from different seed populations. The shaded area indicates the standard error of the mean. 4.5. Scaling to Larger Libraries The preceding experiments evaluate BOBA in settings where exhaustive surrogate inference is still possible, whi… view at source ↗
Figure 6
Figure 6. Figure 6: Scalability of BOBA with library size. Optimization performance of BOBA relative to full-library BO as the library size increases from approximately 105 to 108 molecules. Performance is measured as the area under the top-1000 retrieval curve and normalized by the AUC of the corresponding full-library BO run. All experiments use the ZINC library docked against AmpC, or subsets thereof. scale, we further eva… view at source ↗
Figure 7
Figure 7. Figure 7: Effect of uncertainty estimation on optimization performance. Optimization trajectories for full-library BO and BOBA using Laplace approximation, SWAG, and MC dropout uncertainty estimates. Curves report retrieval of top-ranked candidates as a function of the number of molecules explored across ENAMINE-S-3.9M docked against NEDD4, ENAMINE-S-3.9M docked against CKB, and ENAMINE-5M docked against CKB. Column… view at source ↗
Figure 8
Figure 8. Figure 8: Trajectories of top-1000 and top-10000 molecules recovered for the ENAMINE-5M library (CKB) [PITH_FULL_IMAGE:figures/full_fig_p019_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Trajectories of top-1000 and top-10000 molecules recovered for the ENAMINE-S-3.9M (CKB) library 19 [PITH_FULL_IMAGE:figures/full_fig_p019_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Trajectories of top-1000 and top-10000 molecules recovered for the ENAMINE-S-3.9M (NEDD4) library D. Arm Selection Frequencies To assess whether UCB1 collapses onto a small number of initially high-reward partitions, we analyzed the fraction of selections assigned to each arm over the course of optimization [PITH_FULL_IMAGE:figures/full_fig_p020_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Per-arm selection frequencies for BOBA with UCB1. Each panel reports the fraction of selections assigned to each arm over optimization iterations, with arms color-coded by partition. Experiments are shown for ENAMINE-S-3.9M docked against NEDD4, ENAMINE-S-3.9M docked against CKB, and ENAMINE-5M docked against CKB, across K ∈ {5, 10, 25, 100}. 20 [PITH_FULL_IMAGE:figures/full_fig_p020_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Distribution of docking scores for CKB-5M inside each cluster for different values of K. Outliers not shown. 22 [PITH_FULL_IMAGE:figures/full_fig_p022_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Distribution of docking scores for CKB-3.9M inside each cluster for different values of K. Outliers not shown. 23 [PITH_FULL_IMAGE:figures/full_fig_p023_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Distribution of docking scores for NEDD4-3.9M inside each cluster for different values of K. Outliers not shown. 24 [PITH_FULL_IMAGE:figures/full_fig_p024_14.png] view at source ↗
read the original abstract

Identifying high-utility candidates from massive discrete spaces under expensive evaluations is a recurring challenge across the sciences, with structure-based drug discovery as a prominent example. While surrogate-based optimization can increase sample efficiency by reducing the number of expensive evaluations, modern molecular libraries have reached billions to trillions of compounds, making full-library surrogate inference itself a major computational bottleneck. We introduce BOBa, a bandit-guided surrogate optimization framework that eliminates full-library inference by adaptively allocating computation across partitions of the action space. By treating partitions as arms in a multi-armed bandit, BOBa concentrates inference and evaluations on empirically promising partitions while maintaining principled exploration. Experiments on real-world synthesis-on-demand libraries demonstrate that optimism-under-uncertainty bandits, combined with meaningful action space partitioning, are essential for effective allocation of inference and evaluations. Our findings reveal a tunable tradeoff between screening performance and surrogate inference cost, which supports practical optimization over current libraries, and establishes a viable route to ultra-large library virtual screening.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The paper introduces BOBa, a bandit-guided surrogate optimization framework for identifying high-utility candidates in massive discrete chemical spaces. It treats partitions of the action space as arms in a multi-armed bandit to adaptively allocate surrogate inference and evaluations, avoiding full-library computation. The central claim is that optimism-under-uncertainty bandits combined with meaningful action space partitioning are essential for effective allocation, supported by experiments on real-world synthesis-on-demand libraries that also reveal a tunable tradeoff between screening performance and inference cost.

Significance. If the results hold, the approach could enable practical virtual screening over billion- to trillion-scale libraries by reducing the computational cost of surrogate inference, addressing a key scalability bottleneck in structure-based drug discovery. The explicit tradeoff analysis is a practical contribution.

major comments (2)
  1. [Abstract] Abstract: the claim that 'experiments on real-world synthesis-on-demand libraries demonstrate that optimism-under-uncertainty bandits, combined with meaningful action space partitioning, are essential' is presented without any quantitative results, error bars, baseline comparisons, or details on performance metrics, leaving the central empirical claim without verifiable support in the provided description.
  2. [Experiments (implied by abstract claim)] The load-bearing assumption that partitions admit sufficiently uniform utility within arms (so that partial observations can reliably concentrate on high-utility regions) is not tested against boundary bias. Any fixed partition (e.g., by scaffold or fingerprint clustering) can split chemically similar high-scoring molecules, potentially causing an arm with low observed mean to contain the global optimum; no ablation or sensitivity analysis on partition construction appears to address this.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive feedback. We address each major comment below and indicate where revisions will be made to the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that 'experiments on real-world synthesis-on-demand libraries demonstrate that optimism-under-uncertainty bandits, combined with meaningful action space partitioning, are essential' is presented without any quantitative results, error bars, baseline comparisons, or details on performance metrics, leaving the central empirical claim without verifiable support in the provided description.

    Authors: The abstract is a concise summary; the full quantitative results (performance metrics, baseline comparisons, error bars, and statistical details) appear in the Experiments section. To address the concern, we will revise the abstract to include a brief sentence highlighting key empirical outcomes from the synthesis-on-demand library experiments. revision: yes

  2. Referee: [Experiments (implied by abstract claim)] The load-bearing assumption that partitions admit sufficiently uniform utility within arms (so that partial observations can reliably concentrate on high-utility regions) is not tested against boundary bias. Any fixed partition (e.g., by scaffold or fingerprint clustering) can split chemically similar high-scoring molecules, potentially causing an arm with low observed mean to contain the global optimum; no ablation or sensitivity analysis on partition construction appears to address this.

    Authors: We agree this is an important robustness consideration. The manuscript employs scaffold-based partitioning for chemical interpretability and shows effective allocation in experiments, but does not include an explicit ablation across partition types or direct sensitivity analysis to boundary effects. We will add this analysis in revision, comparing scaffold, fingerprint clustering, and alternative schemes to quantify impact on allocation quality. revision: yes

Circularity Check

0 steps flagged

No significant circularity; applies standard bandits to partitioned spaces without self-referential reductions

full rationale

The manuscript presents BOBa as a framework that treats chemical-space partitions as arms in a multi-armed bandit and allocates surrogate inference accordingly. The central claim rests on experimental demonstration that optimism-under-uncertainty bandits plus meaningful partitioning improve allocation efficiency. No equations, derivations, or self-citations are shown that reduce the reported performance gains to quantities defined by the method itself, to fitted parameters renamed as predictions, or to load-bearing self-citations. The work therefore applies existing bandit algorithms to a new domain without the circular patterns enumerated in the analysis criteria.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides no explicit free parameters, axioms, or invented entities; the method is described at the level of standard bandit concepts applied to partitions.

pith-pipeline@v0.9.1-grok · 5725 in / 1038 out tokens · 18925 ms · 2026-06-26T05:26:05.679583+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

49 extracted references · 42 canonical work pages · 6 internal anchors

  1. [1]

    and Singh, Isha and Levit, Anat and Moroz, Yurii S

    Lyu, Jiankun and Wang, Sheng and Balius, Trent E. and Singh, Isha and Levit, Anat and Moroz, Yurii S. and O’Meara, Matthew J. and Che, Tao and Algaa, Enkhjargal and Tolmachova, Kateryna and Tolmachev, Andrey A. and Shoichet, Brian K. and Roth, Bryan L. and Irwin, John J. , year =. Ultra-large library docking for discovering new chemotypes , volume =. Natu...

    work page doi:10.1038/s41586-019-0917-9

  2. [2]

    and Shakhnovich, Eugene I

    Graff, David E. and Shakhnovich, Eugene I. and Coley, Connor W. , year =. Accelerating high-throughput virtual screening through molecular pool-based active learning , volume =. Chemical Science , publisher =. doi:10.1039/d0sc06805e , number =

    work page doi:10.1039/d0sc06805e

  3. [3]

    A sober look at LLMs for material discovery: are they actually good for bayesian optimization over molecules? , url =

    Kristiadi, Agustinus and Strieth-Kalthoff, Felix and Skreta, Marta and Poupart, Pascal and Aspuru-Guzik, Al\'. A sober look at LLMs for material discovery: are they actually good for bayesian optimization over molecules? , url =. Proceedings of the 41st International Conference on Machine Learning , articleno =. doi:10.48550/arXiv.2402.05015 , year =

    work page doi:10.48550/arxiv.2402.05015

  4. [4]

    and Kumar, Aditya and Gehev, Minko and Aquilanti, Pierre-Yves and Gabb, Henry and Alhossary, Amr and Wagner, Gerhard and Aspuru-Guzik, Alan and Moroz, Yurii S

    Gorgulla, Christoph and Cecchini, Domiziana and Nigam, AkshatKumar and Tang, Ming and Reis, Joana and Koop, Matt and Gottinger, Andrea and Nicoll, Callum Robert and Jayaraj, Abhilash and Cinaroglu, Suleyman Selim and Torner, Ricarda and Seo, Hyuk-Soo and Dhe-Paganon, Sirano and Secker, Christopher and Haddadnia, Mohammad and Malets, Yehor and Hasson, Alex...

    work page doi:10.1101/2023.04.25.537981 2023

  5. [5]

    and Godin, Guillaume and Lehtivarjo, Juuso and Pahl, Axel and Walker, Rachel and Berenger, Francois and Biggs, Jason D

    Landrum, Greg and Tosco, Paolo and Kelley, Brian and Rodriguez-Schmidt, Ricardo and Cosgrove, David and Riniker, Sereina and Gedeck, Peter and Vianello, Riccardo and Schneider, Nadine and Kawashima, Eisuke and N, Dan and Jones, Gareth and Dalke, Andrew and Cole, Brian and Swain, Matt and Turk, Samo and Savelyev, Alexander and Vaucher, Alain and Wójcikowsk...

    work page doi:10.5281/zenodo.8053810

  6. [6]

    Open Babel: An open chemical toolbox , volume =

    O’Boyle, Noel M and Banck, Michael and James, Craig A and Morley, Chris and Vandermeersch, Tim and Hutchison, Geoffrey R , year =. Open Babel: An open chemical toolbox , volume =. Journal of Cheminformatics , publisher =. doi:10.1186/1758-2946-3-33 , number =

    work page doi:10.1186/1758-2946-3-33

  7. [7]

    Uni-Dock: GPU-Accelerated Docking Enables Ultralarge Virtual Screening , volume =

    Yu, Yuejiang and Cai, Chun and Wang, Jiayue and Bo, Zonghua and Zhu, Zhengdan and Zheng, Hang , year =. Uni-Dock: GPU-Accelerated Docking Enables Ultralarge Virtual Screening , volume =. Journal of Chemical Theory and Computation , publisher =. doi:10.1021/acs.jctc.2c01145 , number =

    work page doi:10.1021/acs.jctc.2c01145

  8. [8]

    Proceedings of the 40th International Conference on Machine Learning (ICML 2023) , pages =

    Unifying Molecular and Textual Representations via Multi-task Language Modelling , url =. Proceedings of the 40th International Conference on Machine Learning (ICML 2023) , pages =. 2023 , publisher =. doi:10.48550/arXiv.2301.12586 , author =

    work page doi:10.48550/arxiv.2301.12586 2023

  9. [9]

    Liu , title =

    Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu , title =. Journal of Machine Learning Research , year =

  10. [10]

    Machine Learning , author =

    Auer, Peter and Cesa-Bianchi, Nicolò and Fischer, Paul , year =. Finite-time Analysis of the Multiarmed Bandit Problem , volume =. Machine Learning , publisher =. doi:10.1023/a:1013689704352 , number =

    work page doi:10.1023/a:1013689704352

  11. [11]

    Bandit Algorithms , url =

    Lattimore, Tor and Szepesvári, Csaba , year =. Bandit Algorithms , url =. doi:10.1017/9781108571401 , publisher =

    work page doi:10.1017/9781108571401

  12. [12]

    Proceedings of the 35th International Conference on Neural Information Processing Systems (NeurIPS 2021) , pages =

    Daxberger, Erik and Kristiadi, Agustinus and Immer, Alexander and Eschenhagen, Runa and Bauer, Matthias and Hennig, Philipp , title =. Proceedings of the 35th International Conference on Neural Information Processing Systems (NeurIPS 2021) , pages =. doi:10.48550/arXiv.2106.14806 , year =

    work page doi:10.48550/arxiv.2106.14806 2021

  13. [13]

    Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning

    Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning , author =. Proceedings of the 33rd International Conference on Machine Learning (ICML 2016) , pages =. 2016 , editor =. doi:10.48550/arXiv.1506.02142 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1506.02142 2016

  14. [14]

    and Izmailov, Pavel and Garipov, Timur and Vetrov, Dmitry P

    Maddox, Wesley J. and Izmailov, Pavel and Garipov, Timur and Vetrov, Dmitry P. and Wilson, Andrew Gordon , title =. Proceedings of the 32nd International Conference on Neural Information Processing Systems (NeurIPS 2019) , pages =. 2019 , url =

    2019

  15. [15]

    Kingma and Jimmy Ba , title =

    Diederik P. Kingma and Jimmy Ba , title =. 3rd International Conference on Learning Representations (ICLR 2015) , year =

    2015

  16. [16]

    SGDR: Stochastic Gradient Descent with Warm Restarts

    Ilya Loshchilov and Frank Hutter , title =. 5th International Conference on Learning Representations (ICLR 2017) , year =. doi:10.48550/arXiv.1608.03983 , url =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1608.03983 2017

  17. [17]

    ICML 2023 Workshop on Structured Probabilistic Inference

    Graph Neural Network Powered Bayesian Optimization for Large Molecular Spaces , author=. ICML 2023 Workshop on Structured Probabilistic Inference. 2023 , url=

    2023

  18. [18]

    Chemical space: limits, evolution and modelling of an object bigger than our universal library , volume =

    Restrepo, Guillermo , year =. Chemical space: limits, evolution and modelling of an object bigger than our universal library , volume =. Digital Discovery , publisher =. doi:10.1039/d2dd00030j , number =

    work page doi:10.1039/d2dd00030j

  19. [19]

    The elephant in the lab: synthesizability in generative small-molecule design , volume =

    Papidocha, Sven M and Burger, Andreas and Bernales, Varinia and Aspuru-Guzik, Alán , year =. The elephant in the lab: synthesizability in generative small-molecule design , volume =. doi:10.1016/j.coche.2025.101217 , journal =

    work page doi:10.1016/j.coche.2025.101217 2025

  20. [20]

    , year =

    Shoichet, Brian K. , year =. Virtual screening of chemical libraries , volume =. Nature , publisher =. doi:10.1038/nature03197 , number =

    work page doi:10.1038/nature03197

  21. [21]

    The cost of drug development: A systematic review , volume =

    Morgan, Steve and Grootendorst, Paul and Lexchin, Joel and Cunningham, Colleen and Greyson, Devon , year =. The cost of drug development: A systematic review , volume =. Health Policy , publisher =. doi:10.1016/j.healthpol.2010.12.002 , number =

    work page doi:10.1016/j.healthpol.2010.12.002 2010

  22. [22]

    Active-learning strategies in computer-assisted drug discovery

    Reker, Daniel and Schneider, Gisbert , year =. Active-learning strategies in computer-assisted drug discovery , volume =. Drug Discovery Today , publisher =. doi:10.1016/j.drudis.2014.12.004 , number =

    work page doi:10.1016/j.drudis.2014.12.004 2014

  23. [23]

    Practical considerations for active machine learning in drug discovery

    Reker, Daniel , year =. Practical considerations for active machine learning in drug discovery , volume =. doi:10.1016/j.ddtec.2020.06.001 , journal =

    work page doi:10.1016/j.ddtec.2020.06.001 2020

  24. [24]

    The next level in chemical space navigation: going far beyond enumerable compound libraries , volume =

    Hoffmann, Torsten and Gastreich, Marcus , year =. The next level in chemical space navigation: going far beyond enumerable compound libraries , volume =. Drug Discovery Today , publisher =. doi:10.1016/j.drudis.2019.02.013 , number =

    work page doi:10.1016/j.drudis.2019.02.013 2019

  25. [25]

    and Nicklaus, Marc C

    Warr, Wendy A. and Nicklaus, Marc C. and Nicolaou, Christos A. and Rarey, Matthias , year =. Exploration of Ultralarge Compound Collections for Drug Discovery , volume =. Journal of Chemical Information and Modeling , publisher =. doi:10.1021/acs.jcim.2c00224 , number =

    work page doi:10.1021/acs.jcim.2c00224

  26. [26]

    ICLR 2025 Workshop on Generative and Experimental Perspectives for Biomolecular Design , year=

    Active Learning on Synthons for Molecular Design , author=. ICLR 2025 Workshop on Generative and Experimental Perspectives for Biomolecular Design , year=

    2025

  27. [27]

    Active Learning to Select the Most Suitable Reagents and One-Step Organic Chemistry Reactions for Prioritizing Target-Specific Hits from Ultralarge Chemical Spaces , volume =

    Kozyrev, Vladimir and Sindt, Fran. Active Learning to Select the Most Suitable Reagents and One-Step Organic Chemistry Reactions for Prioritizing Target-Specific Hits from Ultralarge Chemical Spaces , volume =. Journal of Chemical Information and Modeling , publisher =. 2025 , pages =. doi:10.1021/acs.jcim.4c02097 , number =

    work page doi:10.1021/acs.jcim.4c02097 2025

  28. [28]

    Patrick , year =

    Klarich, Kathryn and Goldman, Brian and Kramer, Trevor and Riley, Patrick and Walters, W. Patrick , year =. Thompson Sampling – An Efficient Method for Searching Ultralarge Synthesis on Demand Databases , volume =. Journal of Chemical Information and Modeling , publisher =. doi:10.1021/acs.jcim.3c01790 , number =

    work page doi:10.1021/acs.jcim.3c01790

  29. [29]

    and Gathiaka, Symon and Walters, W

    Zhao, Hongtao and Nittinger, Eva and Yu, Melissa A. and Gathiaka, Symon and Walters, W. Patrick and Tyrchan, Christian , year =. Enhanced Thompson sampling by roulette wheel selection for screening ultralarge combinatorial libraries , volume =. Journal of Cheminformatics , publisher =. doi:10.1186/s13321-025-01105-1 , number =

    work page doi:10.1186/s13321-025-01105-1

  30. [30]

    and Sadybekov, Anastasiia V

    Nazarova, Antonina L. and Sadybekov, Anastasiia V. and Sadybekov, Arman A. and Protopopov, Mykola and Radchenko, Dmytro S. and Moroz, Yurii S. and Tarkhanova, Olga O. and Katritch, Vsevolod , year =. V-synthes2 - the Next Generation Tool for Structure-based Virtual Screening of Giga-scale Chemical Spaces , url =. doi:10.21203/rs.3.rs-7782723/v1 , publisher =

    work page doi:10.21203/rs.3.rs-7782723/v1

  31. [31]

    and Coote, Paul W

    Gorgulla, Christoph and Boeszoermenyi, Andras and Wang, Zi-Fu and Fischer, Patrick D. and Coote, Paul W. and Padmanabha Das, Krishna M. and Malets, Yehor S. and Radchenko, Dmytro S. and Moroz, Yurii S. and Scott, David A. and Fackeldey, Konstantin and Hoffmann, Moritz and Iavniuk, Iryna and Wagner, Gerhard and Arthanari, Haribabu , year =. An open-source ...

    work page doi:10.1038/s41586-020-2117-z

  32. [32]

    Pyzer-Knapp, E. O. , year =. Bayesian optimization for accelerated drug discovery , volume =. IBM Journal of Research and Development , publisher =. doi:10.1147/jrd.2018.2881731 , number =

    work page doi:10.1147/jrd.2018.2881731 2018

  33. [33]

    A Tutorial on Bayesian Optimization

    A Tutorial on Bayesian Optimization , year =. arXiv Preprint , author =. doi:10.48550/arXiv.1807.02811 , url =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1807.02811

  34. [34]

    2023 , publisher =

    Garnett, Roman , title =. 2023 , publisher =

    2023

  35. [35]

    and Shoichet, Brian K

    Lyu, Jiankun and Irwin, John J. and Shoichet, Brian K. , year =. Modeling the expansion of virtual screening libraries , volume =. Nature Chemical Biology , publisher =. doi:10.1038/s41589-022-01234-w , number =

    work page doi:10.1038/s41589-022-01234-w

  36. [36]

    , year =

    Gloriam, David E. , year =. Bigger is better in virtual drug screens , volume =. Nature , publisher =. doi:10.1038/d41586-019-00145-6 , number =

    work page doi:10.1038/d41586-019-00145-6

  37. [37]

    Synthesis Lectures on Artificial Intelligence and Machine Learning

    Settles, Burr , year =. Active Learning , url =. doi:10.1007/978-3-031-01560-1 , journal =

    work page doi:10.1007/978-3-031-01560-1

  38. [38]

    Bayesian Optimal Active Search and Surveying

    Garnett, Roman and Krishnamurthy, Yamuna and Xiong, Xuehan and Schneider, Jeff and Mann, Richard , title =. Proceedings of the 29th International Conference on Machine Learning (ICML 2012) , pages =. 2012 , publisher =. doi:10.48550/arXiv.1206.6406 , url =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1206.6406 2012

  39. [39]

    Introducing the ‘active search’ method for iterative virtual screening , volume =

    Garnett, Roman and G\". Introducing the ‘active search’ method for iterative virtual screening , volume =. Journal of Computer-Aided Molecular Design , publisher =. 2015 , pages =. doi:10.1007/s10822-015-9832-9 , number =

    work page doi:10.1007/s10822-015-9832-9 2015

  40. [40]

    Efficient nonmyopic active search with applications in drug and materials discovery

    Jiang, Shali and Malkomes, Gustavo and Moseley, Benjamin and Garnett, Roman , keywords =. Efficient nonmyopic active search with applications in drug and materials discovery , publisher =. 2018 , copyright =. doi:10.48550/ARXIV.1811.08871 , url =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1811.08871 2018

  41. [41]

    Some aspects of the sequential design of experiments , volume =

    Robbins, Herbert , year =. Some aspects of the sequential design of experiments , volume =. Bulletin of the American Mathematical Society , publisher =. doi:10.1090/s0002-9904-1952-09620-8 , number =

    work page doi:10.1090/s0002-9904-1952-09620-8 1952

  42. [42]

    Ultra-large library screening with an evolutionary algorithm in Rosetta (REvoLd) , volume =

    Eisenhuth, Paul and Liessmann, Fabian and Moretti, Rocco and Meiler, Jens , year =. Ultra-large library screening with an evolutionary algorithm in Rosetta (REvoLd) , volume =. Communications Chemistry , publisher =. doi:10.1038/s42004-025-01758-x , number =

    work page doi:10.1038/s42004-025-01758-x

  43. [43]

    and Sadybekov, Anastasiia V

    Sadybekov, Arman A. and Sadybekov, Anastasiia V. and Liu, Yongfeng and Iliopoulos-Tsoutsouvas, Christos and Huang, Xi-Ping and Pickett, Julie and Houser, Blake and Patel, Nilkanth and Tran, Ngan K. and Tong, Fei and Zvonok, Nikolai and Jain, Manish K. and Savych, Olena and Radchenko, Dmytro S. and Nikas, Spyros P. and Petasis, Nicos A. and Moroz, Yurii S....

    work page doi:10.1038/s41586-021-04220-9

  44. [44]

    Grygorenko, Dmytro S

    Grygorenko, Oleksandr O. and Radchenko, Dmytro S. and Dziuba, Igor and Chuprina, Alexander and Gubina, Kateryna E. and Moroz, Yurii S. , year =. Generating Multibillion Chemical Space of Readily Accessible Screening Compounds , volume =. iScience , publisher =. doi:10.1016/j.isci.2020.101681 , number =

    work page doi:10.1016/j.isci.2020.101681 2020

  45. [45]

    Structural studies of human brain‐type creatine kinase complexed with the ADP–Mg2+NO3-–creatine transition‐state analogue complex , volume =

    Bong, Seoung Min and Moon, Jin Ho and Nam, Ki Hyun and Lee, Ki Seog and Chi, Young Min and Hwang, Kwang Yeon , year =. Structural studies of human brain‐type creatine kinase complexed with the ADP–Mg2+NO3-–creatine transition‐state analogue complex , volume =. FEBS Letters , publisher =. doi:10.1016/j.febslet.2008.10.039 , number =

    work page doi:10.1016/j.febslet.2008.10.039 2008

  46. [46]

    Structure-based design of potent and selective inhibitors of the

    Maspero, Elena and Cappa, Anna and Weber, Janine and Trifirò, Paolo and Amici, Raffaella and Bruno, Agostino and Fagà, Giovanni and Cecatiello, Valentina and Fattori, Raimondo and Leuzzi, Brian and Taibi, Vincenzo and Meroni, Giuseppe and Pasi, Maurizio and Romussi, Alessia and Sartori, Luca and Villa, Manuela and Vultaggio, Stefania and Cirò, Marco and S...

    work page doi:10.1038/s42004-025-01557-4

  47. [47]

    and Vigneron, Seth F

    Liu, Fangyu and Mailhot, Olivier and Glenn, Isabella S. and Vigneron, Seth F. and Bassim, Violla and Xu, Xinyu and Fonseca-Valencia, Karla and Smith, Matthew S. and Radchenko, Dmytro S. and Fraser, James S. and Moroz, Yurii S. and Irwin, John J. and Shoichet, Brian K. , year =. The impact of library size and scale of testing on virtual screening , volume ...

    work page doi:10.1038/s41589-024-01797-w

  48. [48]

    Rotting Bandits

    Levine, Nir and Crammer, Koby and Mannor, Shie , title =. Proceedings of the 31st International Conference on Neural Information Processing Systems (NeurIPS 2017) , pages =. 2017 , publisher =. doi:10.48550/arXiv.1702.07274 , url =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1702.07274 2017

  49. [49]

    and Varun, Begur V

    Naik, Maruti and Raichurkar, Anandkumar and Bandodkar, Balachandra S. and Varun, Begur V. and Bhat, Shantika and Kalkhambkar, Rajesh and Murugan, Kannan and Menon, Rani and Bhat, Jyothi and Paul, Beena and Iyer, Harini and Hussein, Syeed and Tucker, Julie A. and Vogtherr, Martin and Embrey, Kevin J. and McMiken, Helen and Prasad, Swati and Gill, Adrian an...

    2015