by Data MacDermott-Opeskin H, Scheen J, Wognum C, Horton JT, West D, Payne AM, Castellanos MA, Colby S, Griffen E, Cousins D, Stacey J, Reid L, Aschenbrenner JC, Fearon D, Balcomb B, Marples P, Tomlinson CWE, Lithgo R, Godoy AS, Winokan M, Barr H, Lahav N, Lavi M, Duberstein S, Cohen G, Fate G, Lefker B, Robinson R, Szommer T, Lynch N, Minh DDL, La VN T, Kang L, Huddleston K, Renslow R, Tollefson M, Walters WP, Xu C, Hsu J, St-Laurent J, Etsmoberg H, Zhu L, Quirke A, Abdul Haleem MIA, Alibay I, Baid G, Birnbaum B, Bishop KP, Bohorquez H, Bose A, Brown CJ, Burns J, Cai L, Cedeno R, de Cesco S, Chupakhin V, Clark F, Cole DJ, Corbi-Verge C, Danial M, Davi A, Dehaen W, Doering NPD, Dougha A, Dréanic MP, Eakin B, Ehrlich A, Elijosius R, Fülöp J, Gitter A, Goossens K, Gu Y, Head-Gordon T, Hoffer L, Hofmans J, Jiang E, Kaminow B, Khosravi S, Khoualdi AF, Lenselink EB, Liu Z, Liu Y, Liu S, Ma Y, Maher P, Mayer I, Mendez-Lucio O, Mey ASJS, Michel J, Montanari F, Niu T, Ogino R, Palaniappan A, Pan X, Patnaik A, Pham LH, Pinto L, Purnomo J, Rich A, Schaaf L, Schran C, Singh RK, Srilakshmi M, Srivastava SP, Sun K, Sun Z, Talagayev V, Balakrishnan BTS, Titus I, Tkatchenko A, Treyde W, Tricarico G, Tripp A, Vithayapalert N, Wang Y, Wasi AT, Wedig S, Wolber G, Xu B, Zhou W, von Delft F, Lee A, Kirkegaard K, Sjö P, Fraser JS, Chodera JD. Journal of Chemical Information and Modeling 2026. doi: 10.1021/acs.jcim.5c02106
Summary: Computational blind challenges offer an opportunity to assess and accelerate scientific progress. The authors of this manuscript report outcomes and key insights from an open science community blind challenge focused on computational methods in drug discovery, using lead optimization data from the AI-driven Structure-enabled Antiviral Platform Discovery Consortium’s pan-coronavirus antiviral discovery program, in partnership with Polaris and the OpenADMET project. Participants developed and applied computational methods to predict the biochemical potency and ligand poses of small molecules against key coronavirus targets using previously undisclosed data sets as benchmarks. By evaluating submissions across multiple tasks and compounds, the authors assessed methodological strengths, common pitfalls, and areas for improvement. This analysis provides a foundation for best practices in real-world machine learning evaluation.
