Research On Genome-wide Bioassay Model Based On Generalized Principal Component Analysis

Generalized principal component analysis is developed on the basis of principal component analysis,the thoughts of it and principal component analysis are the same,is to replace multiple indexes with fewer indexes to reflect the original index information.For example,there are n samples,p indexes measured by each sample,then there are np data.Because the indexes have an impact on each other,we can find out several comprehensive indexes from p indexes and analyze them,so that we can use less than p indexes to get the results of p indexes.In genome-wide association study,it is an important subject to identify genetic markers associated with disease phenotype,some other clinic or environment factors from a larger number of single nucleotide polymorphisms(SNPs).Against the structure of two-dimensional contingency table of phenotype-SNP,a GPCA model involving all SNPs simultaneously is proposed.Specifically,the matrix of the model is Singular value decomposition(SVD)to reduce the number of parameters in the model,so the model can also be called the Logistic SVD model.From which,a SNP selection criterion is given by describing the distribution difference of the SNP genotypes under different disease phenotype levels.The results of numerical simulations show that the proposed detection criterion is more effective than existing methods.