Artificial Intelligence-Enabled Analysis of WNT Pathway Dysregulation in Bevacizumab-Treated Early-Onset Colorectal Cancer.

Early-onset colorectal cancer (EOCRC) is increasing disproportionately among Hispanic/Latino (H/L) populations and demonstrates substantial molecular and clinical heterogeneity. Although Wingless/Integrated (WNT) pathway alterations are among the most common genomic events in colorectal cancer, their prognostic significance in the context of contemporary systemic therapies, including Bevacizumab, remains incompletely understood. We conducted an integrative clinical-genomic analysis of colorectal cancer cohorts stratified by age at diagnosis, ancestry, and Bevacizumab exposure, interrogating somatic alterations across curated WNT signaling pathway genes. Conversational artificial intelligence agents (AI-HOPE and AI-HOPE-WNT) enabled dynamic cohort construction, treatment-specific subgroup analyses, and pathway-level interrogation through natural language-driven clinical-genomic integration. WNT pathway alterations, predominantly involving APC, were highly prevalent across all cohorts; however, their distribution and clinical associations demonstrated strong treatment-, ancestry-, and age-dependent variability. Bevacizumab-treated tumors exhibited lower mutation frequencies in several WNT regulators, including RNF43, AXIN1/2, TCF7L2, and AMER1, suggesting potential biologic interaction or treatment-related selective pressure. Importantly, WNT pathway alterations were associated with improved overall survival in H/L EOCRC and Non-Hispanic White (NHW) late-onset colorectal cancer, but with worse survival in NHW EOCRC, highlighting distinct ancestry- and age-specific prognostic effects. These findings support the role of the WNT pathway dysregulation as a disparity-aware biomarker framework in colorectal cancer and demonstrate the utility of conversational AI systems for scalable multidimensional clinical-genomic integration in precision oncology.