CACHE #4: Prospective hit-finding tests on the CBL-B kinase-binding domain yield few validated binders
- In this blind, prospective challenge, 23 teams each nominated up to 100 compounds predicted to bind the CBL-B tyrosine kinase binding domain, an immuno-oncology E3 ligase, and the picks were tested experimentally.
- Across more than 1,600 predictions, only a handful from several teams showed any activity; just two were convincingly validated and only one was chemically novel.
- The best workflow combined a binding-similarity classifier, a random forest trained on patent inhibitors and a diffusion-based pose predictor. Authors point to rare hits, weak-ligand retrieval and binding-site dynamics as obstacles.
💡 Points Based on the abstract
Two validated binders out of 1,600+ predictions is a hit rate near 0.1%, and that is the honest number for a target with little chemical matter. Note the winner leaned on patent inhibitors, so the useful lesson is that known-ligand data still beats pure structure-based scoring. Next step for a team: check whether orthogonal assays, not one readout, confirmed the two hits.
📄 Paper · Journal of Chemical Information and Modeling (ACS) ↗ Critical Assessments of Computational Hit-finding Experiments Challenge #4: Targeting the Tyrosine Kinase Binding Domain of the E3 Ligase Casitas B-Lineage Lymphoma Proto-Oncogene B, an Immuno-Oncology Target