This proposal brings together experimental cancer biology and machine-learning genomics to understand how lung cancer cells change identity over time and how those changes help tumors resist treatment. The two collaborating teams will investigate mouse models of small-cell lung cancer (SCLC), a fatal form of lung cancer, across primary tumors, treated tumors, and metastases, generate single-cell RNA and chromatin-accessibility data, and use deep learning to identify candidate regulatory factors that control these shifting cell states. Functional screens, analyzed with new machine learning models, will validate these candidates. By combining the Sage lab’s expertise in cancer models with the Kundaje lab’s expertise in computational analysis and model interpretation, the project aims to pinpoint which molecular switches are most important for keeping cancer cells in therapy-sensitive or therapy-resistant states. The long-term goal is to find new ways to limit tumor plasticity and improve treatment response in this aggressive cancer.