Project II - Leveraging artificial intelligence for integrative multiomics analysis of high-risk populations to understand the biologic etiology of pancreatic cancer

Abstract public domain (NIH RePORTER)

PROJECT SUMMARY – Project 2 Project 2 Pancreatic ductal adenocarcinoma (PDA) is a leading cause of cancer-related mortality in the United States, with rising incidence and death rates. Mortality rates vary across patient populations, and in those at the highest risk of developing PDA, mortality remains elevated even after accounting for established risk factors and socioeconomic measures. This pattern suggests a possible biological contribution. However, high-risk patient populations for which limited data are available have not been extensively investigated in publicly available genomic and clinical databases or in biorepositories, constraining the ability to conduct comprehensive analyses and develop refined precision therapeutic strategies. Advances in artificial intelligence (AI) provide a powerful opportunity to bridge this gap by enabling large-scale, real-world data analyses. To address this gap, we will pursue three specific aims: In Aim 1, we will develop a natural language processing (NLP) model to extract structured and unstructured clinicopathologic data from clinical sources, facilitating standardized and scalable analyses. We will then apply this validated model to characterize the somatic mutational landscape of PDA in Hawai‘i and assess its impact on patient outcomes. In Aim 2, we will analyze clinically derived whole-exome and whole transcriptome sequencing data to evaluate subtype distribution, stromal composition, and stromal activation. Additionally, we will leverage AI-driven spatial transcriptomics to investigate tumor spatial organization, including tumor-stroma interactions, and their impact on patient outcomes. In Aim 3, we will generate patient derived organoids (PDOs) from metastatic PDA biopsies of patients from high-risk populations, addressing a critical lack of data in biorepositories. We will validate these PDOs through genomic, histopathological, and therapeutic sensitivity analyses and correlate findings with clinical outcomes to assess their translational potential. By integrating AI-driven data extraction, molecular characterization, and spatial transcriptomic analysis with patient-derived models, we aim to elucidate the molecular and spatial tumor features underlying PDA disparities and potentially identify novel predictive and prognostic biomarkers to improve patient outcomes.