Prediction of PD-1 immunotherapy (IO) response for KRAS mutated non-small cell lung cancer (NSCLC) based on co-mutations using a computational biological model

Abstract

Background: KRAS mutated NSCLC is a heterogeneous disease due to the impact of co-mutations (co-mut). In preclinical models, KRASco-mut differentially activate downstream pathways and affect the tumor immune microenvironment in diverse ways. Data suggest KRASco-mut may affect sensitivity to PD-1 axis IO. Methods: Genomic information of 2974 NSCLC patients was input into computational biological model (CBM) software (Cellworks Group, San Jose, CA). Computational protein network maps of disease characteristics were generated for each patient. CBM was used to predict sensitivity to PD-1 axis IO in KRASco-mut subsets including: KRAS/TP53, KRAS/CDKN2A, KRAS/STK11, KRAS/KEAP1, KRAS/STK11/KEAP1, KRAS/PIK3CA, and KRAS without co-mut. The 3 key metrics used to predict sensitivity included PD-L1 expression; Dendritic Cell Infiltration Index (9 chemokine markers); and Immunosuppressive Biomarker Expression (14 markers). Correlation of CBM prediction of IO sensitivity was examined in a clinical cohort of 36 KRASmutated NSCLC patients with available overall survival (OS) data from Stanford University treated with PD-1 axis IO during their treatment course. Results: In the overall cohort, withregards to prediction of sensitivity to PD-1 axis IO, CBM predicted the majority of patients with KRAS/KEAP1and KRAS/STK11/KEAP1 to not benefit from IO, whereas CBM predicted the majority of patients with KRAS/ TP53, KRAS/PI3KCA, and KRAS without co-mut to benefit. No definitive predictions could be made for KRAS/STK11and KRAS/CDKN2A. In the clinical cohort of 36 patients treated with PD-1 axis IO during their treatment course, CBM was able to assess 27 of these patients, identifying patients with OS>12 months, with 82.8% positive predictive value, 42.9% negative predictive value, and 75% concordance. Conclusions: CBM predicted certain subsets of KRASmutated NSCLC based on comut are more likely to be sensitive to PD-1 axis IO. In a small clinical cohort of KRASmutated NSCLC treated with PD-1 axis IO, in light of existing biomarkers, CBM identified patients with improved prognosis with good positive predictive value.