Direct detection of genetic dominance from natural variation in human populations. D. Balick1,2, R. Do3, D. Reich3, S. Sunyaev1,2 1) Genetics Division, Brigham and Women's Hospital, Harvard Medical School, Boston, MA; 2) Broad Institute, Cambridge. MA; 3) Department of Genetics, Harvard Medical School, Boston, MA.
Despite ubiquitous evidence for genetic dominance in phenotypic and disease data, direct detection of the dominant or recessive action of natural selection in humans remains elusive. Unlike the analogous question in model organisms, experimental systems designed to differentiate between distinct types of selection in humans are infeasible. As such, there is a need for the development of a statistical test for the mode of selection from natural population samples. Here, we present such a test, derived from the differential action of recessive and dominant selection in populations with distinct demographic histories. We verify the efficacy of this test by showing that genes predicted to be recessive are enriched for genes responsible for recessive disease. Additionally, genes predicted to act dominantly do not show enrichment in this gene set. This pattern is consistent in both a large set of genes associated with autosomal recessive disease, and to a greater extent in a hand curated set of genes with no association with any disease phenotypes in heterozygous form. Aggregating genes in this high quality list into a unit, we find that the statistical test predicts recessivity for this sub-genome, in qualitative agreement with the enrichment analysis. This analysis was repeated for genes known to be responsible for congenital hearing loss, a well-known autosomal recessive disease. We again find agreement between our statistical prediction and the mode of inheritance of the disease phenotype. We confirm the detection of dominant selection by applying this test on the whole genome level to show that derived mutation classes thought to be under biased gene conversion are predicted to be under co-dominant selection. This is consistent with theoretical predictions for conversion of heterozygous sites. Together, these analyses allow us to quantify the joint distribution of selection and dominance effects, both on a per gene and on a whole genome level. This has important implications for medical and population genetics, particularly in helping us to understand the role and relevance of genetic dominance. Additionally, identifying candidate genes under recessive selection may aid in the discovery, understanding, and treatment of diseases with substantial genetic components.
You may contact the first author (during and after the meeting) at