A Framework for Improving the Generalizability of Drug–Target Affinity Prediction Models

• DOI: 10.1089/cmb.2023.0208
• Publication Date: 2023-11-21T18:35:52Z
• Authors: Rıza Özçelik; Alperen Bağ; Berk Atil; Melih Barsbey; Arzucan Özgür; Elif Ozkirimli
• Keywords: Spurious relationship; Data set; Drug target

Statistical models that accurately predict the binding affinity of an input ligand-protein pair can greatly accelerate drug discovery. Such are trained on available interaction data sets, which may contain biases lead predictor to learn set-specific, spurious patterns instead generalizable relationships. This leads prediction performances these drop dramatically for previously unseen biomolecules. Various approaches aim improve model generalizability either have limited applicability or introduce risk degrading overall performance. In this article, we present DebiasedDTA, a novel training framework drug-target (DTA) addresses set such models. DebiasedDTA relies reweighting samples achieve robust generalization, and is thus applicable most DTA Extensive experiments with different biomolecule representations, architectures, sets demonstrate achieves improved in predicting affinities.