Artificial Intelligence-Driven Precision Oncology Uncovers Prognostic Significance of RTK-RAS Alterations in FOLFOX-Treated Early-Onset Colorectal Cancer

Abstract

Abstract The incidence of early-onset colorectal cancer (EOCRC; <50 years) is rising rapidly among populations. Although alterations in the RTK-RAS signaling pathway are central to colorectal cancer (CRC) progression, their prognostic significance in FOLFOX-treated EOCRC remains poorly defined. We analyzed 2,515 CRC cases (H/L = 266; non-Hispanic White [NHW] = 2,249) stratified by ancestry, age at onset, and FOLFOX treatment status using Fisher’s exact, chi-square, and Kaplan-Meier survival analyses. We further employed the AI-HOPE and AI-HOPE-RTK-RAS conversational artificial intelligence platforms to integrate clinical, genomic, and treatment-level data through multi-parameter, natural language-driven queries. Among EO H/L patients, ERBB2 (2.7% vs. 15.4%, p = 0.01) and NF1 (4.1% vs. 19.2%, p = 0.01) mutations were significantly less frequent in FOLFOX-treated compared with untreated patients. Among LO H/L patients, NTRK2 mutations were less enriched in FOLFOX-treated cases (0.0% vs. 6.0%, p = 0.04). In FOLFOX-untreated EO H/L patients, MAPK3 (5.8% vs. 1.0% in EO NHW; p = 0.04) and NF1 (19.2% vs. 6.0% in LO NHW; p = 0.002) mutations were significantly enriched. Among EO NHW patients, IGF1R (2.1% vs. 5.3%, p = 0.04) and ERRFI1 (0.5% vs. 2.6%, p = 0.02) mutations were less frequent in FOLFOX-treated versus untreated cases. In LO NHW patients, multiple RTK-RAS genes (ERBB3, KIT, IGF1R, RET, ALK, FLT3, ERRFI1, ARAF, RAF1) were less enriched with FOLFOX exposure. Survival analyses revealed that RTK-RAS pathway alterations predicted worse overall survival in FOLFOX-untreated EO NHW patients ( p = 0.029) but were associated with improved survival in FOLFOX-treated LO NHW patients ( p = 0.048). Collectively, these findings indicate that RTK-RAS pathway alterations exhibit strong ancestry-, age-, and treatment-specific effects and may serve as precision biomarkers of differential chemotherapy response. The AI-enabled framework markedly accelerated integrative biomarker discovery, supporting its utility for advancing precision oncology in populations affected by EOCRC.

Ähnliche Arbeiten