Enhancing colorectal cancer precision medicine through multi-omics and clinical data integration with artificial intelligence.

Abstract

3603 Background: The integration of multi-omics and clinical data in precision medicine research for colorectal cancer (CRC) is a complex task that requires advanced computational tools. Artificial Intelligence agent for High-Optimization and Precision mEdicine (AI-HOPE) has emerged as a transformative platform, streamlining data integration, analysis, and discovery efforts. AI-HOPE is designed to integrate and analyze multi-omics alongside clinical data, facilitating novel insights into CRC pathogenesis, therapeutic responses, and precision medicine applications. Methods: AI-HOPE leverages Large Language Models (LLMs) to interpret natural language inputs and convert them into executable pipelines for multi-omics and clinical data analysis. Data from The Cancer Genome Atlas (TCGA) and other publicly available CRC datasets were utilized to demonstrate its capabilities. AI-HOPE supports analyses such as identifying mutation and gene expression patterns, pathway enrichment, survival analysis, and therapeutic outcome predictions. Three case studies were conducted: (1) identifying WNT and TGFβ pathway alterations in early-onset CRC (EO CRC) versus late-onset CRC, (2) evaluating the association between specific multi-omics signatures and progression-free survival in patients treated with FOLFOX chemotherapy, and (3) performing a precision medicine query to identify actionable gene mutations and tailored medications for CRC treatment. Results: Overall AI-HOPE answered queries with an accuracy of 0.98 and an F1 score of 0.89 (precision = 0.80). AI-HOPE identified significant alterations in the WNT and TGFβ pathways among EO CRC patients compared to late-onset cases, aligning with findings from published literature. In the second study, the platform revealed that patients with specific transcriptomic signatures (e.g., upregulation of MYC targets) had significantly worse progression-free survival when treated with FOLFOX chemotherapy, further supporting its utility in identifying clinically relevant biomarkers. AI-HOPE was used to query CRC datasets for actionable gene mutations, such as KRAS, BRAF, and MSI-H (microsatellite instability-high), and cross-referenced these findings with drug databases to identify tailored therapies. The analysis highlighted FDA-approved targeted treatments, such as EGFR inhibitors (cetuximab and panitumumab) for KRAS wild-type patients and immune checkpoint inhibitors (pembrolizumab and nivolumab) for MSI-H tumors. AI-HOPE also identified emerging therapeutic options from ongoing clinical trials, showcasing its potential for guiding precision medicine strategies in CRC. Conclusions: This study demonstrates the transformative potential of AI-HOPE in advancing CRC precision medicine research by seamlessly integrating multi-omics and clinical data.