Remodeling of cancer cell architecture by chemotherapy

How chemotherapy reshapes tumor cells-and how these changes influence outcomes such as drug resistance-remains largely unclear. Here, we present a multimodal, global characterization of tumor subcellular organization and its reorganization by chemotherapy. We use self-supervised learning to encode protein coordinates across four orthogonal data modalities: proteome-wide size-exclusion chromatography fractionation (before and after treatment with cisplatin or vorinostat), native-state immunofluorescence imaging, affinity purification, and primary sequence information covering 7,579 proteins. This integrated map resolves 174 subcellular components, spanning molecular assemblies from protein complexes to organelles across a size range of ~10-9 to10-5nm. 58 components undergo significant remodeling upon treatment, recapitulating known mechanisms of action and revealing previously unrecognized alterations in pathways such as cytoskeletal organization and metabolic rewiring. We systematically validate these “chemotherapy-response assemblies” using genome-wide CRISPR knockout drug-sensitivity profiling, identifying which assemblies confer drug sensitivity versus resistance. Chemotherapy-remodeled components serve as convergence points for cancer mutations that predict therapeutic response-including those involved in homologous recombination repair, chromatin remodeling, and double-strand break repair.