Project 1: Racial and ethnic differences in the intra-tumoral microbiome: Impact on colorectal cancer mortality and clinicopathologic correlates

PROJECT SUMMARY/ABSTRACT of Project 1 Colorectal cancer (CRC) is one of the leading causes of cancer death, and it disproportionately impacts African American people whose CRC mortality rate is 40% higher than the US average. Given this and the rising rates of early-onset CRC, particularly among racial and ethnic minority groups, we need new approaches to identify

individuals at higher risk who could benefit from precision screening strategies. Currently, only family history is used to identify individuals who should undergo earlier and more frequent screening; however, over 80% of CRC are diagnosed in people without a family history of CRC, suggesting that we need more comprehensive and

tailored ways to stratify individuals according to their risk. Using polygenic risk scores (PRS), which aggregate all CRC-associated genetic variants into a single score to predict CRC risk, we can identify high risk subgroups from the general population who are currently considered average-risk but who would likely benefit from targeted

prevention and screening interventions. However, arguably one of the most critical ethical and scientific challenges related to this is the fact that current PRS are substantially more effective in predicting risk in non- Hispanic White individuals due to the consequences of the Euro-centric bias present in genetic research. A

promising strategy to address this is the use of functional genomic data to guide PRS development, which we propose in Aim 1. We will derive functional genomic scores from single cell data for chromatin accessibility and gene-expression for each cell type and state relevant to CRC (Aim 1a). In Aim 1b we will use machine learning

methods to incorporate the functional genomic scores to build and validate PRS based on data from the largest and most racially and ethnically diverse genetic epidemiologic consortium (123,000 CRC cases and 228,000 controls, including 45,000 cases and 121,000 controls who do not identify as non-Hispanic White). As an increasing

number of companies are offering PRS direct to consumers, we urgently need to evaluate the clinical value of PRS for risk stratified CRC screening. Therefore, we will use the microsimulation model MISCAN to quantify the distributional cost-effectiveness of PRS for risk stratification and estimate the optimal screening strategy in terms

of equity and efficiency based on the PRS across a racially and ethnically diverse population (Aim 2). Our team is ideally positioned to lead this effort given we discovered most of the common genetic risk factors for CRC, have a strong track record conducting single cell multi-omic data analyses, developed comprehensive risk

prediction models using cutting edge machine learning approaches, and have a successful history of working with community advisors. To promote the translation of a novel PRS that performs well across different racial and ethnic groups, we deliberately focus on increasing diversity in genetic studies of CRC and will make data

and our models publicly available. As PRS are increasingly entering clinical care, our findings will address key cancer disparities as we seek to improve CRC prevention by guiding personalizing interventions that will reduce the burden of CRC across racial and ethnic groups.