Abstract 5788: Molecular profiling of NSCLC patients associates CT-based radiomic features with a tumor microenvironment favorable for immunotherapy

Abstract Radiomics provides a powerful, noninvasive way to capture tumor heterogeneity and predict therapy response. However, its clinical utility is limited by feature interpretation. Here, we integrate radiomics with tumor transcriptomic profiles to enhance biological interpretation of CT-based features linked to an immune-responsive tumor microenvironment (TME), in non-small cell lung cancer (NSCLC) treated with acasunlimab, a DuoBody® PD-L1×4-1BB bispecific antibody, +/- pembrolizumab. This study used contrast-enhanced CT images (n=62) and gene expression (HTG EdgeSeq) from tumor biopsies (n=25) of advanced NSCLC subjects in GCT1046-01 (NCT03917381) and GCT1046-04 (NCT05117242). Gene signatures (n=104) linked to cancer and immune pathways were analyzed. Radiomic features were extracted using Picture Health Px Platform, creating a 4-feature composite radiomics score optimized to assess clinical benefit (one quantitative vessel tortuosity (QVT) and three textural features (Laws, Gabor, and GLCM). Performance and treatment specificity were cross-validated and evaluated in 85 NSCLC patients treated with chemotherapy. The radiomics score was not associated with baseline BMI, ECOG score, gender, age, or tumor size. Clustering radiomics features with mRNA signatures identified distinct biological pathways, such as immune regulation, angiogenesis and apoptosis. Specifically, QVT was negatively associated with immune regulation and antigen presentation, while positively correlated with hypoxia and tumor immune evasion. In contrast, GLCM correlated with B cell signaling, immune checkpoint activation and lymphocyte infiltration, indicating a TME suitable for immunotherapy. Laws showed positive correlation with antiviral defense and immune activation, and negative association with tumor immune evasion. Gabor was negatively correlated with cancer associated fibrosis and extracellular matrix remodeling. These results show that integrating radiomics with tumoral mRNA data identified key characteristics of a TME conducive for immunotherapy, advancing the interpretability of AI-derived imaging features. The composite radiomics score was independent of traditional prognostic factors and reflected immunotherapy-specific mechanisms, highlighting its potential as a noninvasive tool for characterizing tumor biology and determining immunotherapy relevance in NSCLC patients. *equal author contribution, LKB, AH Citation Format: Lauren K. Brady, Amogh Hiremath, Mohammed Qutaish, Kai Zhang, Kaustav Bera, Anant Madabhushi, Sri Sridhar, Nora Pencheva, Patricia Garrido-Castro, Maria Jure-Kunkel, Omar Jabado, Matthew Loya, Erin Henitz, Regina Aquino, Teng Jin Ong, Nathaniel Braman, Brandon W. Higgs. Molecular profiling of NSCLC patients associates CT-based radiomic features with a tumor microenvironment favorable for immunotherapy [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 5788.