Multiomics data integration methods to discover putative causal variants, genes and patient heterogeneity for Alzheimers disease

PROJECT SUMMARY Despite the success of genome-wide association studies (GWAS) in identifying over 70 susceptibility loci for Late-onset (LO) Alzheimer’s disease (AD), AD related disease and endophenotypes, it remains challenging to pinpoint 1) which are truly causal AD variants; 2) the molecular processes that cause AD; and 3) how AD patients are pathogenically different from each other. Emerging resources for the study of AD genetics, including sequence, functional genomics and epigenomic data, provide unparalleled opportunity to investigate these questions at different molecular levels. We propose a multiomics data integration project to characterizes AD risk for both genetic variants and individual patients, by developing and applying a series of novel computational approaches using Bayesian hierarchical modeling, variable selection and multivariate analysis, for analyses of a wide range of existing and novel AD multiomics data. These methods are designed to integrate many genetic factors — single nucleotide variants, brain tissue molecular traits such as gene expression, alternative splicing, alternative polyadenylation, methylation, histone acylation and proteomics, and various functional annotations for coding and non-coding regions — into a coherent framework for discovery of causal AD variants and genes, and understand patient heterogeneity. Our goals are to 1) combine genetic association evidence from population and family-based studies of diverse ancestry backgrounds; 2) incorporate functional information to infer putative causal genetic variants; 3) identify novel molecular traits and QTLs for alternative polyadenylation and differentially methylated regions in brain tissues; 4) dissect AD association signals using multiple molecular traits across a comprehensive collection of brain tissues and relevant cell types; and 5) characterize AD patients’ risk profiles using causal effects at different molecular levels across brain tissues. Our methods and bioinformatics analyses will be engineered into a high-quality toolbox to also facilitate multiomics studies of other complex diseases. We will develop fine-mapping methods for family and multi-ancestry data, integrated with thousands of genomic functional annotations, to identify putative causal variants from whole-genome sequences. We will develop a new method to generate alternative polyadenylation from RNA-seq data in brain tissues of AD patients and controls, and fine-map its QTL. We will develop and apply new approaches to fine-map differentially methylated regions in brains, to colocalize QTLs for dozens of molecular traits with AD, and to identify novel gene-level associations using predicted molecular traits. Causal effects estimated at variants and gene levels will be integrated to identify new AD gene-sets and pathways, and to characterize risk profiles for AD patients. Causal variants and genes discovered from our project will provide insight for development of therapeutic drugs targeting at specific cellar processes. The multiomics risk profiles built for AD patients will improve clinical trial designs for AD drug development, paving the path to personalized therapeutics.