Alzheimer's disease (AD) is a devastating disorder that causes relentlessly progressive loss of memory and cognition. AD affects 5.4 million people in the United States and its cost to society is estimated at $180 billion annually. Current treatment options offer some symptomatic benefit, but do not alter the course of this disorder. Thus, there is a great interest in uncovering novel genetic AD-risk loci that can potentially lead to new drug treatment targets as well as identify individuals at early risk for AD for such future drug studies. Several genome-wide association studies (GWAS) of AD have been conducted to search for risk loci, with many such studies collecting a broad range of phenotypic, transcriptomic, and proteomic data both to enhance gene detection as well as gain molecular insight into mapped genes. However, existing well-powered GWAS have had limited success in identifying such risk loci. Here, we propose novel quantitative tools to improve the performance of gene mapping for AD that leverages two important observations from the multitude of GWAS analyses of AD performed to date. First, the genetic origins for AD involve potentially thousands (or even tens of thousands) of trait loci. Second, family-based estimates of heritability for AD tend to be ~50% greater than heritability estimated from GWAS SNP data. While larger GWAS samples may fill this heritability gap, we argue a substantial portion of genetic variance for AD is likely due to non-additive effects that include higher order interactions. Combining existing knowledge that a large number of loci affect AD with the concept that a substantial portion of genetic variance in AD liability is due to higher-order interactions, we propose novel quantitative tools to improve our understanding of the genetic basis of this debilitating disorder. We will explore plausible quantitative-genetics models of the non-additive contributions to genetic variance as well as create novel tests for detecting interactions that jointly analyze multiple AD-related phenotypes together for improved performance.
In Aim 1, we will explore plausible genetic models that have potential to yield significant differences in AD heritability estimates between family-based and GWAS approaches due to higher-order interactions (Aim 1a) and then, under such models, develop a multi-phenotype LD score regression that allows for testing such interactions in GWAS data (Aim 1b).
In Aim 2, we propose novel tools for gene mapping in GWAS that leverage multiple AD-related phenotypes to identify specific SNPs across the genome that interact with other (latent) factors to help explain higher-order interaction effects (Aim 2a). We further will scan for stereotypical outcomes of those interactions on related phenotypes, including studying correlated changes in gene or protein expression within the brain as a function of disease status or other variables of interest (Aim 2b). Upon creating and validating these quantitative tools, we will apply them to various GWAS studies of Alzheimer's disease possessing SNP, transcriptomic, and proteomic data to identify and validate novel AD risk loci and further implement these tools in user-friendly software for public use in other disorders (Aim 3).
This project will develop novel quantitative models and tools for studying interaction effects among genetic variables within genomewide association studies of Alzheimer's disease (AD) that possess SNP, transcriptomic, and proteomic data. Using these models, we hope to discover novel risk loci for AD that could improve our knowledge of the genetic underpinnings of this disease and further provide valuable insight for novel drug treatment targets. We will also create and distribute public software implementing these tools to enable their wide application to a variety of other genetic studies of complex disorders.
