UCLA

An ever increasing wealth of biological data has become available in recent years, and with it, the potential to understand complex traits and extract disease relevant information from these many forms of data through computational methods. Understanding the genetic architecture behind complex traits can help us understand disease risk and adverse drug reactions, and to guide the development of treatment strategies. Many variants identified by genome-wide association studies (GWAS) have been found to affect multiple traits, either directly or through shared pathways. Analyzing multiple traits at once can increase power to detect shared variant effects from publicly available GWAS summary statistics. Use of multiple traits may also improve accuracy when estimating variant effects, which can be used in polygenic scores to stratify individuals by disease risk. This dissertation presents a method, CONFIT, for combining GWAS in multiple traits for variant discovery, and explores a few potential multi-trait methods for estimating polygenic scores. Computational methods can also be used to identify patients already suffering from disease who would benefit from treatment. Towards this end, this dissertation also presents work on deep learning to detect patients with orbital disease from image data with high accuracy and recall.