Artificial intelligence–enabled analysis of prognostic PI3K pathway alterations in African American colorectal cancer patients treated with FOLFOX.

Abstract

e15511 Background: The phosphoinositide 3-kinase (PI3K) signaling pathway regulates tumor growth, metabolism, and treatment response in colorectal cancer (CRC). African American (AA) patients experience a disproportionate CRC burden yet remain underrepresented in molecular studies assessing chemotherapy-associated biomarkers. The prognostic significance of PI3K pathway alterations across ancestry, age at diagnosis, and FOLFOX exposure remains poorly defined. Methods: We performed a retrospective analysis of 2,562 CRC patients with genomic, clinical, and treatment data from public datasets. Cohorts were stratified by ancestry (AA vs. non-Hispanic White [NHW]), age at diagnosis ( < 50 vs. ≥50 years), and FOLFOX treatment. PI3K pathway alterations (including MTOR, AKT3, and PPP2R1A) were analyzed using Fisher’s exact test. Overall survival (OS) was assessed by Kaplan–Meier and Cox models. AI-HOPE conversational AI agents enabled reproducible cohort stratification and integrated analyses. Results: FOLFOX-exposed AA late-onset CRC (LOCRC) tumors showed fewer mTOR mutations than unexposed tumors (0% vs. 9.9%, p = 0.04). In NHW early-onset CRC (EOCRC), FOLFOX exposure was associated with lower AKT3 mutation frequency (0.8% vs. 3.0%, p = 0.04). NHW LOCRC tumors treated with FOLFOX had reduced PPP2R1A (1.7% vs. 3.5%, p = 0.03) and mTOR mutations (4.5% vs. 7.8%, p = 0.007). PI3K pathway alterations were associated with improved OS in AA LOCRC not treated with FOLFOX (p = 0.039), NHW EOCRC treated with FOLFOX (p = 0.0008), and NHW LOCRC not treated with FOLFOX (p = 0.005). However, PI3K pathway alterations did not have prognostic significance among FOLFOX-treated or -untreated AA EOCRC patients and among FOLFOX-treated AA LOCRC patients. Conclusions: PI3K pathway alterations demonstrate ancestry-, age-, and treatment-specific prognostic associations in CRC. AI-enabled, ancestry-aware analyses highlight the PI3K pathway as a context-dependent biomarker and support the integration of artificial intelligence to advance equitable precision oncology.