13 / 14
Page 46
conferenceseries
.com
Volume 11
Journal of Proteomics & Bioinformatics Open Access
Computational Biology 2018
September 05-06, 2018
September 05-06, 2018 Tokyo, Japan
International Conference on
Computational Biology and Bioinformatics
J Proteomics Bioinform 2018, Volume 11
DOI: 10.4172/0974-276X-C1-113
A robust statistical method to integrate genotype and gene expression data identifies novel associated loci
Sarmistha Das and Indranil Mukhopadhyay
Indian Statistical Institute, India
G
enome wide association studies identify many SNPs that are associated with disease traits. However, single marker test
might miss SNPs with moderate effect. Moreover, gene expression contains information about the deregulation of genes
when compared between cases and controls. Due to multiple testing or other issues some signals may remain unidentified
especially when the sample size is not too large. Moreover, RNA being unstable than DNA, high cost is involved in RNA analysis
leading to smaller sample for expression data than genotype data. No standard statistical procedure is available that integrates
data from various sources to decode biologically sound interpretation on heritable traits. This motivates us to propose a novel
method that tests for multi-loci association in the existing scenario. Based on a two-stage regression method our method
essentially concatenates genome-wide expression data and disease-associated SNP data, when sample size for expression data
is much smaller than genotype data. We integrated the information contained in both data sources into a latent variable based
model. Our simple yet powerful multi-loci association test integrates two databases that broadcasts more of the deep-seated
features comprehensively in a single test, which might be lost when datasets are considered in singularity. We also developed
asymptotic distribution of our test statistic for fast calculation of p-value for real data set. Extensive simulation confirms that our
method is powerful and robust to many genetic models. We have received promising result and identified few novel markers at
genome-wide level even with a small gene expression dataset related to psoriasis.
sarmisthadascu@gmail.com