AI–Physics Convergence for Early Discovery: From Tractability & Large-Scale Screening to Candidate Prioritization
- Demonstrate a reproducible workflow that integrates tractability/pocket mapping, large-scale ligand-based and structure-based virtual screening, potency ML using interaction fingerprints, and physics-based scoring (MM-PBSA, RBFE/FEP+, dissociation FE)
- Share applications across two target studies. The first being, 8M-scale screening, pharmacophore/LBVS+SBVS funnel to 496 candidates. The second being, ensemble docking, ML selection, physics-based rescoring
- Demonstrate the translation of computational signals into ranked, purchase-ready lists and model-informed designs in LiveDesign
- Highlight the advantages and caveats in speed and accuracy (e.g., HMR for throughput, cycle-closure curation to remove high-uncertainty RBFE transformations, off-target filters like hERG). Experimental validation is still ongoing