AbstractPopulation structure in genotype data has been extensively studied, and is revealed by looking at the principal components of the genotype matrix. However, no similar analysis of population structure in gene expression data has been conducted, in part because a naïve principal components analysis of the gene expression matrix does not cluster by population. We identify a linear projection that reveals population structure in gene expression data. Our approach relies on the coupling of the principal components of genotype to the principal components of gene expression via canonical correlation analysis. Futhermore, we analyze the variance of each gene within the projection matrix to determine which genes significantly influence the projection. We identify thousands of significant genes, and show that a number of the top genes have been implicated in diseases that disproportionately impact African Americans.Author SummaryHigh dimensional, multi-modal genomics datasets are becoming increasingly common, which warrants investigation into analysis techniques that can reveal structure in the data without over-fitting. Here, we show that the coupling of principal component analysis to canonical correlation analysis offers an efficient approach to exploratory analysis of this kind of data. We apply this method to the GEUVADIS dataset of genotype and gene expression values of European and Yoruban individuals, finding as-of-yet unstudied population structure in the gene expression values. Moreover, many of the top genes identified by our method have been previously implicated in diseases that disproportionately impact African Americans.
Correctness results for on-line robust principal components analysis
