DSAI-01 ENHANCING CAR T CELL SPECIFICITY IN LUNG CANCER BRAIN METASTASES THROUGH LOGIC-GATED TARGET DISCOVERY USING SINGLE-CELL TRANSCRIPTOMICS

Abstract Selecting effective and specific cell surface targets is critical for developing CAR T cells against lung cancer brain metastases. Traditionally, CAR T cell targets for metastatic disease are selected based on markers from the primary tumor site, neglecting potential antigenic drift and differences between primary and metastatic tumors. To address this limitation, we developed a logic-gated combinatorial framework leveraging single-cell RNA sequencing (scRNA-seq) data to systematically identify optimal pairs of cell surface markers. This method integrates scRNA-seq datasets from healthy tissues (Tabula Sapiens), primary non-small cell lung cancer (NSCLC), and NSCLC brain metastases, forming a comprehensive atlas for precise antigen selection tailored to metastatic lesions. By employing Boolean logic gates (AND, OR, NOT) across varying expression thresholds, we prioritized marker combinations that maximize tumor coverage while minimizing normal tissue targeting, specifically within the brain microenvironment. Using this framework, we used machine learning to select 100 candidate genes and quantified their specificity and coverage. Our approach not only highlighted known CAR T cell targets but also revealed novel, previously unexplored targets. For instance, while MUC1 is a traditional CAR target, covering 56.6% of primary NSCLC tumor cells with 9.7% normal tissue expression, pairing MUC1 with SDK1 via an OR gate increased tumor coverage to 76.6%, maintaining low normal cell coverage (10.8%). Importantly, the analysis identified EFNA5, a novel target yet to be explored in CAR T cell therapy, which alone covered 68.2% of brain metastatic tumor cells with minimal normal expression (6%). Combining EFNA5 with CDH1 (OR gate) further enhanced tumor coverage to 82.9%, maintaining normal cell coverage around 10%. Our results identified distinct antigen combinations optimal for primary versus metastatic brain lesions, highlighting the importance of tumor-microenvironment-specific target selection. This data-driven strategy offers a robust framework to design selective and effective CAR T cell therapies against lung cancer brain metastases.