ClinOracle: Hierarchical AI Prediction of Target Binding and Patient-Derived Functional Activity Across Diverse Therapeutic Targets

The platform's key innovation is conditioning functional activity predictions on predicted target engagement, which better reflects the biological pathway from binding to cellular response. In benchmarks across five therapeutic targets spanning oncology, ClinOracle demonstrated improved predictions of patient-derived functional activity compared to models that treat binding and activity independently. This approach could help pharmaceutical researchers filter compound libraries earlier in the pipeline, potentially reducing late-stage clinical failures. The study is published as a preprint on bioRxiv and has not yet undergone peer review.