Single-cell multiomic drug response profiling of PRISM-multiplexed cancer cell lines sequenced with SBX

Inter-individual variability in drug response is a key challenge in precision oncology. Despite recent breakthroughs in targeting RAS-mutant cancers, clinical responses remain heterogeneous. To systematically dissect the molecular determinants of drug sensitivity and resistance, we implemented a high-throughput single-cell multi-omic profiling framework using a PRISM pool: a barcoded co-culture platform of ∼400 human cancer cell lines spanning diverse lineages and genetic backgrounds. We treated the PRISM pool with two RAS inhibitors (BI-2865 and RMC-6236), as well as a negative control (DMSO) and a positive control (Panobinostat), and collected cells at early time points (3h and 12h) to capture initial drug response dynamics. We employed a modified 10x Genomics Flex protocol enabling simultaneous capture of the whole transcriptome, a targeted proteome (~320-plex Proteintech panel), and PRISM identity via expressed DNA barcodes. The prototype Sequencing by Expansion (SBX) platform was leveraged to generate ~230 billion reads, enabling the analysis of 315,547 high-quality single cells. Drug treatments induce diverse and cell line-specific shifts in transcriptional and proteomic states. Applying Non-negative Matrix Factorization (NMF) on each cell line's single-cell transcriptome data and then performing hierarchical clustering, we identified 20 drug-responsive, recurrently co-expressed gene meta-programs (MPs), condensing 1831 underlying programs related to cell cycle, growth, structure, and stress response. Multiple MP dynamics are associated with drug sensitivity and often stratified by genetic background. Notably, cell lines that exhibited a delayed stress-response MP (non-responsive at 3 hours but catching up at 12 hours) exhibited high drug sensitivity. To identify key protein players in drug response, we conducted differential expression (DE) analysis for each cell line comparing drug treatments and controls, identifying recurrently significant DE proteins whose altered expression was also significantly associated with cell viability. Importantly, these protein changes frequently lacked a corresponding significant change in the expression of their encoding RNA transcripts, underscoring the power of multi-modal profiling to more comprehensively illuminate functional mechanisms driving drug response. Our study establishes a scalable paradigm for linking genotype, transcriptome, and proteome to pharmacologic phenotype at single-cell resolution across genetically diverse human models. These data, enabled with massively high-throughput sequencing using SBX, provide a rich resource for mechanistic discovery and rational design of combination therapies targeting the RAS pathway.