Abstract 3369: Identifying biomarkers of resistance for immunotherapy-treated patients with non-small-cell lung cancer using real-world evidence and machine learning

Abstract Immunotherapy (IO) has revolutionized the management of non-small-cell lung cancer (NSCLC), leading to groundbreaking improvements in patient survival. However, resistance to IO in some patients remains a challenge. Existing IO resistance biomarkers are based on either clinical or genomic factors. Combining genomic factors with patient health characteristics is pivotal in developing new resistance markers and improving the accuracy of predictions of treatment resistance. The objective of this study is to develop a data-driven model with no a priori assumptions using real-world evidence (RWE) data to identify predictors of treatment resistance in NSCLC patients treated with first-line IO. By integrating clinical and genomic features, we aim to assess the differences in the importance of genomic mutations under various clinical conditions and identify new biomarkers associated with IO resistance. Using clinical and genomic features from real-world evidence (RWE) Flatiron dataset for an NSCLC cohort treated with first-line IO, we developed a machine learning (ML) model to identify predictors of treatment resistance. Using progression-free survival (PFS) as an endpoint, the model identified key contributing factors, a combination of clinical and genomic features, where the importance of genomic mutations varies under different clinical conditions. Specifically, in patients with ECOG scores of 0 or 1 and no metastasis, the CDKN2A mutation was associated with reduced PFS; for those with ECOG scores of 0 or 1 and with metastasis, CDKN2A was associated with decreased PFS and KRAS with improved PFS; and for patients with ECOG scores of 2-4, KRAS was associated with improved PFS. Our results underscore the value of RWE in identifying resistance biomarkers, demonstrate the efficacy of a data-driven methodology able to extract combinations of important features without prior assumptions, and highlighted the importance of integrating clinical and molecular data to understand patient-level risk factors of treatment resistance. Using our methodology, we pinpointed statistically significant biomarkers linked to an increased risk of treatment resistance in NSCLC patients undergoing first-line IO treatment. Citation Format: Chao Fang, Natasha Markuzon, Sajan Khosla. Identifying biomarkers of resistance for immunotherapy-treated patients with non-small-cell lung cancer using real-world evidence and machine learning [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 3369.