Compound–protein interaction prediction is essential for drug discovery, yet current models struggle with accuracy due to reliance on sequence-based data and limited labeled datasets. Here, the ...