Evaluating Cancer Drug-Targeting Pathways with Large Language Models

Integrating artificial intelligence into drug discovery holds great potential for prioritizing therapeutic targets. This study presents a novel framework combining bioinformatics tools and AI-driven evaluations to streamline target identification. Using pancreatic adenocarcinoma (PAAD) as a case study, we analyzed 252 candidate genes and associated pathways derived from the PAGER database. ChatGPT-4o evaluated pathways by scoring them across seven categories using structured prompts and weighted criteria to ensure robustness. Our approach demonstrated a statistically significant differentiation between PAAD-related and unrelated pathways (t(489)=-12.06, p<0.00001, Hedges’ g=1.24).Topranking pathways included the pancreatic cancer and pancreatic adenocarcinoma pathways. Candidate genes were ranked using normalized pathway significance and genespecific contributions, combined into a weighted formula. This approach highlighted key targets including AURKB, POLA1, and RRM2. These findings highlight the potential of generative AI to automate and accelerate target discovery, offering an adaptable methodology for diverse therapeutic areas.