A GPU program to compute SNP-SNP interactions in genome-wide association studies

With the recent advances in the next generation sequencing technologies, short read sequences of human genome are made more accessible. Paired end sequencing of short reads is currently the most sensitive method for detecting somatic mutations that arise during tumor development. In this study, a novel approach to optimize the detection of structural variants using a new short read alignment program is presented. Pairwise interaction effects of the Single Nucleotide Polymorphisms (SNPs) have proven to uncover the underlying complex disease traits. Computing the disease risk based on the interaction effects of SNPs on a case - control study is a difficult problem. As another part of the thesis, a fast GPU program that can calculate the chi-square statistics of SNP-SNP interactions and output the significant interacting SNPs is presented. The algorithm is applied to the datasets of seven common diseases obtained from Wellcome Trust Case Control Consortium (WTCCC). The algorithm computed the significant SNP pairs much faster than the existing algorithms and also identifies 3 significant pairs associated with genes IL23R and C11orf30 which are associated with pathogenesis in the Crohns disease dataset.

Name thesis:degree_name

Master of Science in Bioinformatics - (M.S.)

Discipline thesis:degree_discipline

Computer Science

Year

2013

Author dc:creator

• Ramakrishnan, Srividya

Contributors dc:contributor

• Usman W. Roshan
• Zhi Wei
• Alexandros V. Gerbessiotis

• Human genome
• Short read sequences
• Short read alignment
• Bioinformatics
• Computer Sciences

Repository record dc:identifier

https://digitalcommons.njit.edu/theses/289

OAI identifier oai:identifier

oai:digitalcommons.njit.edu:theses-1288