Predictive prioritization of enhancers associated with pancreatic disease risk.

Authors

Wang L, Baek S, Prasad G, Wildenthal J, Guo K, Sturgill D, Truongvo T, Char E, Pegoraro G, McKinnon K, Pancreatic Cancer Cohort Consortium, Pancreatic Cancer Case-Control Consortium, Hoskins JW, Amundadottir LT, Arda HE

Affiliations

• Laboratory of Receptor Biology and Gene Expression, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892, USA.
• Laboratory of Receptor Biology and Gene Expression, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892, USA; Laboratory of Translational Genomics, Division of Cancer Epidemiology & Genetics, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892, USA.
• Laboratory of Translational Genomics, Division of Cancer Epidemiology & Genetics, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892, USA.
• Vaccine Branch, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892, USA.
• Laboratory of Receptor Biology and Gene Expression, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892, USA. Electronic address: efsun.arda@nih.gov.

Abstract

Genetic and epigenetic variation in enhancers is associated with disease susceptibility; however, linking enhancers to target genes and predicting enhancer dysfunction remain challenging. We mapped enhancer-promoter interactions in human pancreas using 3D chromatin assays across 28 donors and five cell types. Using a network approach, we parsed these interactions into enhancer-promoter tree models, enabling quantitative, genome-wide analysis of enhancer connectivity. A machine learning algorithm built on these trees estimated enhancer contributions to cell-type-specific gene expression. To test predictions, we perturbed enhancers in primary human pancreas cells with CRISPR interference and quantified effects at single-cell resolution using RNA fluorescence in situ hybridization (FISH) and high-throughput imaging. Tree models also annotated germline risk variants linked to pancreatic disorders, connecting them to candidate target genes. For pancreatic ductal adenocarcinoma risk, acinar regulatory elements showed greater variant enrichment, challenging the ductal cell-of-origin view. Together, these datasets and models provide a resource for studying pancreatic disease genetics.