CircRNA binding-site prediction stops reading only the RNA: CPRISP encodes the protein too, reaching mean AUC 0.9874 across 37 benchmarks
Prediction of circRNA-RBP binding sites has leaned on transcript features while making limited use of protein-specific information. CPRISP feeds circRNA sequence composition, physicochemical properties, secondary structure and pretrained embeddings together with global and local protein-sequence representations into a multi-scale residual classifier. Across 37 benchmark datasets it reaches a mean AUC of 0.9874 versus 0.9230 for CircSSNN, and holds 0.9861 under sequence-similarity-controlled partitions.