Fine scale mapping of genetic loci associated with human disease

Genome-wide association studies (GWAS) use custom SNP arrays that provide ef- fective genetic coverage, but do not detail the association of confirmed susceptibil- ity regions with as dense a marker map as possible. In this thesis I will describe a fine mapping study across 14 associated regions, in 8,000 samples from three dis- eases (Type 2 diabetes mellitus (T2D), coronary artery disease (CAD) and Graves Disease (GD)) and a control group, including over 5,500 successfully genotyped SNPs. We defined using Bayes theorem sets of SNPs (credible sets) that were 95% likely (posterior probability) to contain the causal disease variants. In three of the 14 regions TCF7L2 and CDKN2A/B in T2D and CTLA4 in GD, we found that much of the posterior probability after the fine mapping rested on a single SNP. In seven of the 14 regions (CDKN2A/B in CAD, CTLA4 in GD, and CDKAL1, FTO, HHEX, TCF7L2) and CDKN2A/B in T2D), including the three just mentioned, the credible sets are relatively small. For these regions, the fine mapping experi- ment has provided useful information, at least in excluding large numbers of SNPs from being causal. Almost none of the SNPs in our credible regions had obvious functions, illustrating our lack of knowledge of genome sequence in modulating gene expression and susceptibility to common disease. Based on this experiment, I outline and discuss the possibilities and challenges in identifying causal variants for complex traits through fine mapping of GWA signals. Further, I evaluate the possibility of using genotype imputation in the context of fine mapping, both as a method for fine mapping and as a way to increase efficiency in fine mapping studies.