SAPDTA: A Novel Protein Segment Capture Strategy for Drug-Target Affinity Prediction

Predicting drug-target affinity (DTA) is becoming increasingly vital in the field of drug discovery. Currently, many methods focus solely on the overall encoding of proteins, overlooking the abundant information contained within protein peptides. Therefore, this paper proposes a novel protein segment capture strategy for drug-target affinity prediction (SAPDTA), which is designed to extract local protein features through a local block capture approach. This strategy supports adaptive segmentation of amino acid chains, enabling more flexible extraction of protein structure information at different levels. A hybrid dual-network bilinear interaction module is proposed to address the challenge of protein feature extraction at various scales. Moreover, bilinear interaction blocks are employed to combine and process the chemical properties of drugs with the biological characteristics of their targets. SAPDTA’s performance is assessed using two publicly accessible DTA datasets (Davis and KIBA). According to the experimental results, SAPDTA demonstrates competitive performance compared to existing models across all evaluation metrics. Furthermore, visualization results on the ToxCast dataset highlight the model’s sensitivity to complex drug structures, revealing its capability to understand underlying structure-function relationships.