Journal: bioRxiv

Article Title: Overcoming the accuracy–generalization tradeoff in docking and scoring for prospective virtual screening

doi: 10.64898/2026.08.03.742480

Figure Lengend Snippet: Clusters of IL17A experimentally validated hits.

Article Snippet: LY3509754 (MedChemExpress, HY-139206) was used as a positive-control IL17A inhibitor.

Techniques:

Journal: bioRxiv

Article Title: Overcoming the accuracy–generalization tradeoff in docking and scoring for prospective virtual screening

doi: 10.64898/2026.08.03.742480

Figure Lengend Snippet: Prospective virtual screening campaigns were conducted against four therapeutically relevant targets representing distinct target classes and binding-site architectures. CD73 (nucleotidase), IRAK4 (kinase), and Factor XI (FXIa protease) were targeted through orthosteric inhibition, whereas IL-17A, a protein–protein interaction (PPI) target, was targeted through an allosteric binding site. For each target, the experimentally determined hit rate, a representative validated hit, its biochemical potency ( IC 50), and its structural similarity to the closest previously reported inhibitor (ECFP4 Tanimoto coefficient) are shown. The prospective campaigns identified 56 active compounds from 183 tested for CD73, 19 from 127 for IRAK4, 13 from 299 for Factor XI, and 6 from 194 for IL-17A. Across all four targets, representative hits exhibited low structural similarity to known inhibitors (ECFP4 Tanimoto coefficients of 0.22–0.27), demonstrating the ability of the virtual screening workflow to prospectively identify chemically novel scaffolds across both conventional orthosteric targets and more challenging protein–protein interaction targets.

Article Snippet: Zimlovisertib (MedChemExpress, HY-19836) was used as a positive-control IRAK4 inhibitor.

Techniques: Binding Assay, Inhibition

Journal: bioRxiv

Article Title: Overcoming the accuracy–generalization tradeoff in docking and scoring for prospective virtual screening

doi: 10.64898/2026.08.03.742480

Figure Lengend Snippet: (a) Prospective virtual screening workflow. Approximately 80 billion compounds from Enamine REAL Space were computationally screened and progressively prioritized to 153 compounds for procurement, of which 127 were experimentally evaluated. The number of compounds remaining after each filtering stage and the corresponding biochemical and cellular hit rates are shown. (b) Characterization of validated hits. Top, single-concentration biochemical inhibition measured at 100 and 10 μM . Middle, molecular weight (MW) and topological polar surface area (TPSA) distributions of validated hits. Bottom left, UMAP projection of ECFP4 fingerprints comparing validated hits (orange) with previously reported IRAK4 inhibitors (blue). Bottom right, distribution of ECFP4 Tanimoto similarity between validated hits and the closest known IRAK4 inhibitor, illustrating the chemical novelty of the identified compounds. (c) Representative validated IRAK4 inhibitors. Chemical structures, biochemical and cellular IC 50 values, physicochemical properties, and representative dose-response curves for two non-nucleotide inhibitors.

Article Snippet: Zimlovisertib (MedChemExpress, HY-19836) was used as a positive-control IRAK4 inhibitor.

Techniques: Concentration Assay, Inhibition, Molecular Weight