|
Abstract:
|
Understanding the influence of human genetics on microbiome composition offers both a better understanding of how genetics influences down-stream traits as well as the innate component of the microbiome. However, identification of genetic variants related to microbial beta-diversity poses a grand challenge due to the high dimensionality of both data types, inherent structure in the data, need for control for population stratification and relatedness, and limited sample sizes. To boost power, we propose to conduct multi-SNP analysis wherein the cumulative effect of a group of SNPs (e.g. in a gene) on beta diversity is assessed. To capture structure, we embed both the SNPs and microbiome data inside of kernels and using the kernel RV, a generalized measure of correlation. A finite sample distribution facilitates analytic p-value computation and we propose extensions to allow for adjustment of the top principal components of genetic variability as well as family structure and relatedness. We apply the approach to identify genetic variants associated with vaginal microbiome composition.
|
