Two Stage High Dimensional Statistical Analysis Under Genetic Model Uncertainty

In the typical genome-wide association analysis(GWAS),it is necessary to identify tens of thousands or even millions of single SNP markers,and then screen out the genetic effects associated with complex diseases.But the price of this method is too high,and the resource utilization rate is low.Two stage analysis is a common method in genome association analysis,which makes the detection process more convenient and saves the cost.In addition,most of the existing methods construct test statistics by assuming the known genetic model.If the model is mistakenly specified,it may lead to the reduction of power.Therefore,the regression equation can be constructed to solve this problem when the genetic model is uncertain.In this paper,a two-stage high-dimensional statistical method based on genome-wide analysis is proposed in the case of genetic model uncertainty.In the first stage,max F was used to screen significant gene loci.The unimportant gene loci were eliminated for preliminary screening;In the second stage,the genetic model is introduced to select variables,and the non gene information is considered in the model,and the nonlinear additive model is established to further improve the effect of gene association analysis.Then,the selected gene loci are brought into the uncertain regression equation of the genetic model,and a variety of penalties are used to select the variables of the model,and the parameter estimates are given,and the genetic model is identified.Finally,through the analysis of the actual simulation results,the fitting effect is very good.