Uncovering Causal Protein Markers to Characterize Pancreatic Cancer Etiology and Improve Risk Prediction

Pancreatic cancer (PC) is the third leading cause of cancer mortality in the United States. The vast majority of PC are pancreatic ductal adenocarcinoma (PDAC). The etiology of PDAC is not fully understood. Basic research supports a crucial role of certain proteins in PDAC development. Epidemiological studies also have identified

multiple candidate protein biomarkers for PDAC. However, findings with many of these protein biomarkers have been inconsistent, potentially due to major methodological limitations, such as selection bias and uncontrolled confounding. Besides understanding etiology, identifying causal protein biomarkers can potentially contribute to

improving risk prediction. For PC, substantial efforts have been made to identify high-risk populations for improving PC screening. However, the performance of available PC risk prediction models remains unsatisfactory. There are critical needs to 1) apply a novel study design with reduced limitations of conventional

biomarker studies for characterizing PC causally related protein biomarkers to improve the etiology understanding; and 2) develop improved prediction models that may effectively facilitate PC risk assessment. One strategy to potentially decrease limitations of unmeasured confounding is to use genetic instruments for

assessing the relationship between proteins and PDAC. While our previous studies have utilized proteins measured in blood, it is also critical to study pancreatic ductal tissue, the most relevant tissue for PDAC development, as levels of many proteins show tissue-specific effects. The proposed project will apply a series of

new studies to address these important knowledge gaps. Specifically, we will 1) conduct a study to identify putative causal protein biomarkers for PDAC risk by applying novel methods (Aim 1); 2) functionally characterize top protein biomarkers for their roles in PC biology (Aim 2); and 3) develop and validate prediction models for

PC risk, by incorporating newly identified candidate protein biomarkers and integrating results from multiple statistical methods (Aim 3). Given the strong pilot data, unique resources, and our team's extensive expertise and experience, we are uniquely positioned to conduct this project. Our study will generate important new

knowledge for PC etiology, and develop improved PC risk prediction models.