Lies and Liabilities: Computational Assessment of High-Throughput Screening Hits to Identify Artifact Compounds

Hits from high-throughput screening (HTS) of chemical libraries are often false positives due to their interference with the assay detection technology. In response to this problem, we generated the largest publicly available library of chemical liabilities and developed a free webtool to predict HTS artifacts, both of which are described herein. More specifically, we generated, curated, and integrated HTS datasets for thiol reactivity, redox activity, and luciferase (firefly and nano) activity. Using these curated datasets, we developed and validated several Quantitative Structure-Interference Relationship (QSIR) models to predict nuisance behavior. The resulting models showed balanced accuracy (BA) for external datasets ranging from 62 to 78%. These models were employed for virtual profiling of the National Center for Advancing Translational Science’s (NCATS) in-house library, and 256 compounds for each assay were selected for the experimental validation of nuisance behavior. The BA for these external predictions ranged from 58 to 78% for compounds within the applicability domains of the models. Our findings suggest that the QSIR models developed and validated herein identify nuisance compounds among experimental hits more reliably than popular PAINS filters. The models developed in this study, along with aggregation models previously developed by our group (SCAM Detective) were implemented in “Liability Predictor,” an online tool which may be used as part of chemical library design or for triaging HTS hits. Both the models and the curated datasets are publicly available at https://liability.mml.unc.edu/.