Statistical Methods for Recovering GWAS Data

Genome-wide association studies (GWAS) are used to investigate associations between genetic variants and health and disease. GWAS often use a “chip” to genotype single nucleotide polymorphisms (SNPs) spanning the genome and test for association between genotype and phenotype at each SNP. The topic of this dissertation is methods for recovering some of the markers that are typically discarded or not analyzed in a GWAS. During data cleaning, prior to the statistical analysis, many genetic markers are discarded because they fail to meet standard quality control criteria. In addition, some analysis results are filtered out because they are considered unreliable. However, there are some discarded data that could be recovered and used in the analysis. For instance, markers that fail to meet a cutoff p-value for the Hardy-Weinberg equilibrium (HWE) test are typically discarded, as are markers with minor allele frequency below some arbitrary cutoff. In addition, markers on the X-chromosome are often not analyzed because sex chromosome analyses are not as straightforward as autosomal analyses, and the statistical methods for testing association on X-chromosome markers are not well established or well tested. In order to make use of more information from any given GWAS, the standard quality control criteria could be modified and more flexible and data specific analysis methods could be developed. This should have the potential to increase power.\ud Genetic variation and environmental/behavioral factors interact to cause almost all human diseases with great public health significance. Refinements in data analysis will help improve the process of identifying the genetic factors and their interactions. These are important in order to provide better prevention and treatment for human diseases so to maintain public health. The methods in this dissertation are proposed and investigated to help with providing a better public health.\u