Escaping the drug-bias trap: using debiasing design to improve interpretability and generalization of drug-target interaction prediction

Considering the high cost associated with determining reaction affinities through in-vitro experiments, virtual screening of potential drugs bound with specific protein pockets from vast compounds is critical in AI-assisted drug discovery. Deep-leaning approaches have been proposed for Drug-Target Interaction (DTI) prediction. However, they have shown overestimated accuracy because of the drug-bias trap, a challenge that results from excessive reliance on the drug branch in the traditional drug-protein dual-branch network approach. This casts doubt on the interpretability and generalizability of existing Drug-Target Interaction (DTI) models. Therefore, we introduce UdanDTI, an innovative deep-learning architecture designed specifically for predicting drug-protein interactions. UdanDTI applies an unbalanced dual-branch system and an attentive aggregation module to enhance interpretability from a biological perspective. Across various public datasets, UdanDTI demonstrates outstanding performance, outperforming state-of-the-art models under in-domain, cross-domain, and structural interpretability settings. Notably, it demonstrates exceptional accuracy in predicting drug responses of two crucial subgroups of Epidermal Growth Factor Receptor (EGFR) mutations associated with non-small cell lung cancer, consistent with experimental results. Meanwhile, UdanDTI could complement the advanced molecular docking software DiffDock. The codes and datasets of UdanDTI are available at https://github.com/CQ-zhang-2016/UdanDTI.