GENOMEWIDE ANALYSIS OF INTERACTIONS IN ALZHEIMER DISEASE USING NEURAL NETWORKS

Vanderbilt University, Nashville TN

Investigators

Linked publications & trials

Abstract

DESCRIPTION (provided by applicant): Alzheimer Disease (AD) is a devastating neurodegenerative disorder and is the most common cause of dementia in the elderly. Much evidence suggests that AD clearly has a measurable genetic component. Many gene-mapping studies have found genes associated with AD but most of them have failed to replicate in follow-up studies. Even the confirmed genetic factors only explain a small amount of AD heritability, leaving most of the AD due to genetic factors unexplained. One reason for this Is that most studies on AD (and other complex diseases) only examine a single genetic variant at a time for association to AD. It Is l^nown, however, that gene-gene and gene-environment Interactions are ubiquitous, and likely play a role in human disease. This has been documented in model organisms, and when investigated properly, has been found to influence disease In humans. Yet most studies do not investigate Interactions among genetic and environmental factors that Influence AD risl<because few computational methods exist that can fully tal^e advantage of the wealth of data generated In contemporary genome-wide association studies (GWAS). Here we propose developing two major Improvements to an existing machine learning method (Grammatical Evolution Neural Networl<s;GENN) that already has the ability to detect interactions that Influence disease risk In genome-wide scale data. One extension will improve the model fitting capabilities of GENN while reducing computation time required by the algorithm. The second extension will allow researchers to incorporate Into the analysis existing biological knowledge from publicly accessible databases. This will guide the search for Important disease-Influencing interactions to biologically plausible solutions. We will confirm that these extensions work properly and measure the power and efficiency gains with realistic simulation studies. These extensions will be implemented in a freely available software package that will be accessible to investigators researching genetics of other complex diseases. We will apply the GENN method to a GWAS dataset with approximately 500 AD cases and 500 healthy age-matched unaffected Individuals to discover gene-gene and gene-environment Interactions that Influence AD risk. Finally, we will access publicly available GWAS data from other published studies to validate our findings. In this study we propose developing two novel major extensions to an existing statistical method that is capable discovering genetic factors that cause human disease. We will use the improved version of this methodology to discover genetic and environmental causes of Alzheimer Disease. We will confirm our findings using independent data.