Pathway Based Genome-wide Association Studies In CUDA Platform

Genome-wide Association Study is an effective method by which can determine the association with inheritance and a phenotype or certain diseases in range of the whole genome research.It can successfully screen the variant sites associated with complex diseases or phenotype from within the scope of human genome sequence variation.However,Genome-wide Association Study at present is based on a statistically significant meaning to identify genetic variation which contains its limitations.Pathway-based Genome-wide Association Study is able to consider the interaction in the gene or genes associated multiple mutation loci,for better exploiting and mining genome-wide data.It is often considered to a supplement of genome-wide association studies which only focus on the most significant sites.In this paper,the authors proposed solutions for the following two problems:1).The computational efficiency of existing algorithms are low and these algorithms are based on the CPU serial algorithms;2).The current evaluation methods for GWAS studies are strictly depending on the distribution of genetic data.Firstly,CUDA platform is proposed in this paper to establish a model for full genetic pathway-based analysis,The algorithm is based on Pathways of Distinction Analysis,by making full use of GPU high parallelism ability to significantly improve the time efficiency of pathway analysis.By analysis,the fine-grained data parallel algorithm is faster than serial data algorithm to an order of magnitude.Besides,we implement our model replying on the CUDA platform and adopt a part of full genetic data as experiment data.The experiment results present that our model will improve the computational efficiency on the basis of ensuring the accuracy of pathway analysis,and the speed-up ratio reaches up 120.In addition,this paper puts forward a method to evaluate all the genetic pathways which is based on principal component analysis(PCA),PCA is an effective system assessment method,is widely used in decision analysis.It is mapped to many indexes by using the dimension reduction above a few indicators,and can quantitatively measure the original parameters on the contribution of each principal component.In this paper,the whole genetic pathways is viewed as a system,each pathway as a specific variable,and quantified each pathway by using the distance metric,ultimately through principal component analysis method to evaluate the pathways which are associated with a phenotype or disease.The method bypasses the defect of high sample distribution requirements which is common in existing pathway analysis methods.And the final experiment results show that the application works.