A New Pipeline Of Genome-wide Association Study And Its Application On Autism

Carrying a case-control based analysis on a large dataset, Genome-wide association study (GWAS) scans the whole genome and is able to detect the genetic marker associated with a certain phenotype (disease). GWAS has been proved to be a powerful approach in studying polygenic inheritance disorder. However, an efficient GWAS needs adequate training samples and effective modeling methods, and thus remains a challenge in computational biology. In this paper we present a new GWAS pipeline capable of integrating various sources of SNP microarray datasets, which saves the burden of collecting adequate samples. By applying property machine learning algorithms, our pipeline can narrow down the candidate SNPs as genetic markers with high prediction accuracy.Autism is a typical polygenic inheritance disorder. Polygenic inheritance disorder is a phenotypic characteristic that is attributable to two or more genes and each genes play a slightly role, which makes GWAS suitable for this kind of study. Together the mechanism for autism is still uncertain. By applying our pipeline on public datasets which contains 837 samples (400 training samples, 437 testing samples), 500 568 SNPs, a profile based on random forest including 37 SNPs was obtained with high classification accuracy on both bootstrap and independent validation. One SNP, three genes and seven cytobands related to the profile were confirmed in previous studies. In sum, our pipeline can integrate various sources of SNP microarray datasets, and could output a profile associated with a certain phenotype with high accuracy. The profile obtained can be used as a clue for the biologists. All of these make it suitable for application in the future.