Vinsamlegast notið þetta auðkenni þegar þið vitnið til verksins eða tengið í það: http://hdl.handle.net/1946/29539
There is great interest in analyzing multiple phenotypes simultaneously to increase power, to increase power of genome-wide association studies (GWAS). This is due to lack of power in individual studies and the fact that many studies share a genetic basis. Here we design and implement an approach for analyzing multiple phenotypes simultaneously, using O'Brien's method of combining phenotypes and LD score regression to adjust for genomic inflation.
A genomic correction is needed to account for inflation in the test statistics, and the traditional genomic correction factor tends to be an overestimate. We therefore provide a proof that LD score regression is a valid approach to use with O'Brien's combination method. This method distinguishes between confounding and polygenicity thereby reducing overestimation in genomic inflation correction. We wrote a program to combine phenotypes according to the developed method. We test it using simulations and showed that it can increase power. We applied it to cancer phenotypes, and discovered novel loci, as well as an increase in power for a known cancer locus.
En route to the main objective we examined stepping stones required for the combination method. We computed an Icelandic LD score map and show that the average Icelandic LD score is higher than from a map computed from the 1000 genomes consortium (1000G). We estimate heritability of a set of phenotypes and show that the estimates are upwardly biased if the 1000G map is used to estimate heritability in an Icelandic cohort. We show that there is significant genetic correlation between a cluster of those phenotypes.