A transferable perturbation proteomics framework for phenotype- 1 based annotation and antiproliferative potency scoring of antineoplastic agents

Abstract

Although antineoplastic agents target diverse proteins, whether they converge on conserved cellular response states remains unclear. Here, we generate a TMTpro 18-plex quantitative proteomics atlas of 367 antineoplastic agents in HepG2 cells and identify recurrent drug response phenotypes (DRPs). The higher-order organization of DRP is conserved across seven independent proteomic datasets and shows limited concordance with chemical structure and annotated mechanisms of action. Projection of 189 external profiles representing 94 compounds achieves 42.7% balanced reference-DRP recovery, a 3.8-fold enrichment above the permutation-derived baseline. DRPs also encode antiproliferative potency, enabling derivation of a six-protein Composite Proteomic Score (CPS-6; TBL1X, NAPA, PLEKHA7, ACSL1, ARL8B, and CCDC90B) that retains associations with relative antiproliferative potency across independent proteomic and transcriptomic datasets. Proteomic similarity further reveals phenotypic convergence among mechanistically distinct compounds and enables mechanism annotation of external compounds and natural products. An interactive portal enables projection of new perturbation proteomes for phenotype-based annotation and antiproliferative potency estimation. Together, our study establishes a transferable proteomic framework for antineoplastic agents, providing a scalable resource for mechanistic hypothesis generation, drug repositioning, and phenotype-guided drug discovery.