Pooled single-cell screening in colorectal cancer identifies transcriptional modules of clinical relevance unlocked by oncogenes

While oncogenic mutations shape colorectal cancer biology and therapy response, their prognostic value remains low. Cluster-based classification of patient cancer transcriptomes has shown greater promise for prognosis, yet these systems do not account for the roles of oncogenes in establishing cancer phenotypes. Here, we create and validate a prognostic classifier for colorectal cancer based on transcriptional programs induced by oncogenes. To systematically investigate oncogenic drivers, we employed a barcoded library of colorectal cancer-associated oncogene variants across a panel of genetically diverse colorectal cancer cell lines. We profiled the transcriptomes of over 100,000 transgenic cells and used machine learning to define transcriptional modules capturing key functional traits. Our analysis revealed heterogeneity on the cell-to-cell level and context-dependent gene expression patterns induced by oncogenes. We identified overarching gene expression modules reflecting core oncogenic processes, including cancer cell plasticity, inflammatory response, replicative stress, and epithelial-to-mesenchymal transition. These modules enabled a functional classification that linked oncogenic signalling states to distinct transcriptional profiles. We demonstrated their prognostic value by stratifying clinical colorectal cancer cohorts into high- and low-risk groups. Although partially correlated with established clinical parameters, the modules provided additional prognostic information, improving survival prediction and therapy stratification beyond existing classification systems. In summary, our study establishes a framework that connects oncogenic mutations to core transcriptional modules. By integrating experimental models with clinical data, we provide a resource for investigating colorectal cancer progression and oncogene-specific vulnerabilities, facilitating future research and precision oncology.