Deep Learning Reveals Mechanisms of Resistance to CDK4/6 Inhibitors

This study uses a novel deep learning model to uncover mechanisms of response and resistance to palbociclib, a cyclin-dependent kinase 4 and 6 inhibitor (CDK4/6i) revolutionizing breast cancer treatment. By mapping multiprotein assemblies across 90 genes, the model identified eight core assemblies that predict sensitivity or resistance to palbociclib more accurately than single-gene biomarkers. These assemblies span pathways related to cell-cycle control, growth factor signaling, and chromatin regulation. Notably, a histone regulatory complex involving KAT6A, TBL1XR1, and RUNX1 was found to drive resistance by promoting S-phase entry. Validated through CRISPR–Cas9 experiments and patient-derived xenografts, this approach provides a comprehensive framework for linking genetic profiles to CDK4/6i resistance, offering new insights for personalized cancer therapy.