To better understand drug resistance mechanisms to CDK4/6 inhibitors and inform precision medicine, we analyze real-world multi-omics data from 400 HR+/HER2- metastatic breast cancer patients treated with CDK4/6 inhibitors plus endocrine therapies, including 200 pre-treatment and 227 post-progression samples. The prevalences of ESR1 and RB1 alterations significantly increase in post-progression samples. Integrative clustering analysis identifies three subgroups harboring different resistance mechanisms: ER driven, ER co-driven and ER independent. The ER independent subgroup, growing from 5% pre-treatment to 21% post-progression, is characterized by down-regulated estrogen signaling and enrichment of resistance markers including TP53 mutations, CCNE1 over-expression and Her2/Basal subtypes. Trajectory inference analyses identify a pseudotime variable strongly correlated with ER independence and disease progression; and revealed bifurcated evolutionary trajectories for ER-independent vs. ER-dependent drug resistance mechanisms. Machine learning models predict therapeutic dependency on ESR1 and CDK4 among ER-dependent tumors and CDK2 dependency among ER-independent tumors, confirmed by experimental validation.

VIEW THE PUBLICATION