Submission
| Title: | Exploring Novel Insights into Alcohol Use Disorder and related traits: A Functional Polygenic Score Approach |
| Presenter: | Melissa Robins |
| Institution: | Indiana University School of Medicine |
| Authors: | MM Robins1,2, AB Chen1,2, X Yu2, X Chu2, Z Zhang1, J Huang1, N Green1,2, H Gao1,2, X Xuei2, JL Reiter1,2, Y Wang1,2, HJ Edenberg2,3, D Lai2 and Y Liu1,2 1Center for Computational Biology and Bioinformatics, 2Department of Medical and Molecular Genetics, and 3Department of Biochemistry and Molecular Biology, Indiana University School of Medicine, Indianapolis, IN 46202, USA |
Abstract
| Background/Significance/Rationale: | Polygenic Scores (PGS) have evolved into powerful tools for assessing genetic predisposition to complex traits. Recent studies suggest that integrating functional information to select SNPs to calculate PGS (referred to as functional PGS) both enhances predictive accuracy and improves portability across populations. In this study, we use a functional PGS to uncover mechanisms underlying Alcohol Use Disorder (AUD) and related traits. |
| Methods: | We developed an innovative approach to construct functional PGS by integrating a diverse range of multi-modal functional genomics data, including the effects of genetic variants on gene expression measured in Massively Parallel Reporter Assays (MPRA) in neuronal and glial cells, chromosomal conformation data, and chromatin accessibility data derived from single-cell multiome analyses of postmortem brain tissues. |
| Results/Findings: | We designed MPRA to evaluate the impacts of 23,232 variants (11,564 enhancer and 7,936 3’-untranslated region variants) on gene expression in human neuroblastoma (SH-SY5Y), oligodendroglioma (HOG), and astrocytoma (CCF-STTG1) cell lines. These variants are associated with a broad spectrum of substance use disorders including problematic alcohol use, opioid use disorder and cannabis use disorder. Our results suggest that 25% of the enhancer and 22% of the 3’UTR variants impact gene expression in at least one cell type. These functional variants were used to train a machine learning model to predict the potential impacts of variants that have not been measured. Using the functional variants identified in our analysis and an interpretable machine learning framework, we built PGS for AUD. |
| Conclusions/Discussion: | Our analyses point to key genes and pathways related to neuroinflammation, responses to oxidative stress and neurodegeneration. Integrating functional data into PGS analysis can go beyond risk assessment and enhance our understanding of genetic mechanisms affecting AUD and related traits. |
| Translational/Human Health Impact: | Our results also suggest that functional PGS can provide greater insight into the heterogeneity of AUD risk and lead to better treatment personalization. |
