Date of Award
Spring 2013
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Bioinformatics
First Advisor
Merrill, Stephen
Second Advisor
Bozdag, Serdar
Third Advisor
Valdar, William
Abstract
Type 2 diabetes results from both genes and the environment. Mapping genetic loci in animal models can help identify genes that are involved in type 2 diabetes to better understand the disease. Heterogeneous stock (HS) rats are derived from eight inbred founder strains and maintained in a breeding strategy that minimizes inbreeding. HS rats have a highly recombinant genome, which allows for rapid fine-mapping of complex traits genome-wide. However, this results in a complicated set of relationships between animals that is non-existent in traditional genetic mapping methods. To fine-map traits involved in type 2 diabetes, multiple diabetic phenotypes were collected in 1,038 HS male rats and these animals were genotyped using the Affymetrix 10K SNP array. Following ancestral haplotype reconstruction, a mixed modeling approach was used to identify genetic loci involved in two phenotypes suggestive of diabetes: fasting glucose and glucose area under the curve after a glucose tolerance test. Sibship was used as a random effect in the model to account for the complex family relationships. A genome-wide significant marker interval was detected on chromosome 11 for fasting glucose with a 95% confidence interval of 5.75 Mb. Genome-wide significant marker intervals were also detected on chromosomes 1,3, 10, and 13 for glucose area under the curve, with the average 95% confidence interval for these loci being only 3.15 Mb. A multilocus modeling technique involving resample model averaging was applied to the fasting glucose phenotype. This technique determines how frequently each locus is detected when resampling a portion of the original data-set, thus reducing potential false positives. Multilocus modeling results for fasting glucose coincided with the significant marker interval demonstrated in the mixed modeling approach. Both approaches are effective at detecting significant marker intervals that are expected to be involved in the phenotype of interest with a greater resolution over traditional methods.