SPH Calendar

What: "Extended Homozygosity Score Tests to Detect Positive Selection in Genome-wide Scans"

When: 3:30 p.m. Wednesday, Oct 15

Where: 5-125 Moos T

Ruzong Fan, from the Department of Statistics at Texas A&M University, will discuss "Extended Homozygosity Score Tests to Detect Positive Selection in Genome-wide Scans," during the seminar. A social tea will be held at 3 p.m. in A434 Mayo. All are welcome to attend.

Abstract
One of the surest signatures of recent positive selection is a local elevation of linkage disequilibrium. Detecting such hitchhiking effects is difficult because haplotypes are not directly observable. Fortunately, extended stretches of homozygosity serve as a surrogate indicator of recent positive selection. In this article, we develop test statistics to detect excess homozygosity. We consider three tests: (a) an extended genotype-based homozygosity (EGH) test, (b) a hidden Markov model (HMM), and (c) an extended haplotype-based homozygosity (EHH) test. The null hypothesis of all three tests assume random mating and Hardy-Weinberg equilibrium. They differ in how they treat linkage disequilibrium. The EGH test assumes linkage equilibrium and gives too many false positives. The EHH test conditions on existing linkage disequilibrium and it tests Hardy-Weinberg equilibrium of haplotypes. The HMM test stands between these two extremes and assumes pairwise but not higher-order disequilibrium interactions. We evaluate by simulation the false positive rates for the EGH and HMM tests under the null hypothesis of EGH and find that HMM is more conservative. All three methods are then applied to the HapMap Phase II data. We are able to replicate previous findings of strong positive selection in the 11 genomic regions SLC19A2, EDAR, LCT, PDE11A, SLC30A9, PCDH15, SLC24A5, HERC1, CHST5/ADAT1/KARS, LARGE, and BCAS3. Over the entire genome, our statistics score lowest in Africans and highest in east Asians, with Europeans intermediate. This pattern is consistent with the accepted ages of these populations. Across the genome, EHH scores are generally low with sharp spikes in only a few regions. Thus, it appears that Hardy-Weinberg equilibrium is valid for most of the genome. EGH and HMM tend to be much higher, most likely because they fail to fully capture linkage disequilibrium. The plots of the statistics suggest regions besides the 11 just mentioned that are candidates for recent positive selection. Further detailed examination of these additional regions is warranted.

Questions? Call 612-624-4655