Sparse regression models for unraveling group and individual associations in eQTL mapping.

Wei Cheng,Xiang Zhang,Wei Wang,Yu Shi

doi:10.1186/s12859-016-0986-9

Abstract

BackgroundAs a promising tool for dissecting the genetic basis of common diseases, expression quantitative trait loci (eQTL) study has attracted increasing research interest. Traditional eQTL methods focus on testing the associations between individual single-nucleotide polymorphisms (SNPs) and gene expression traits. A major drawback of this approach is that it cannot model the joint effect of a set of SNPs on a set of genes, which may correspond to biological pathways.ResultsTo alleviate this limitation, in this paper, we propose geQTL, a sparse regression method that can detect both group-wise and individual associations between SNPs and expression traits. geQTL can also correct the effects of potential confounders. Our method employs computationally efficient technique, thus it is able to fulfill large scale studies. Moreover, our method can automatically infer the proper number of group-wise associations. We perform extensive experiments on both simulated datasets and yeast datasets to demonstrate the effectiveness and efficiency of the proposed method. The results show that geQTL can effectively detect both individual and group-wise signals and outperforms the state-of-the-arts by a large margin.ConclusionsThis paper well illustrates that decoupling individual and group-wise associations for association mapping is able to improve eQTL mapping accuracy, and inferring individual and group-wise associations.Electronic supplementary materialThe online version of this article (doi:10.1186/s12859-016-0986-9) contains supplementary material, which is available to authorized users.

Highlights

As a promising tool for dissecting the genetic basis of common diseases, expression quantitative trait loci study has attracted increasing research interest
Expression quantitative trait loci mapping aims at identifying single nucleotide polymorphisms (SNPs) that influence the expression level of genes
We propose two improved models, geQTL+ and geQTL-ridge, which optimize the search for significant individual associations, which is the main computational bottleneck of the algorithm

Summary

Results

We perform extensive experimental study using both simulated and real eQTL datasets to evaluate the performance of our methods. We observe that by decoupling individual and group-wise associations, the proposed models (geQTL, geQTL+, and geQTL-ridge) are more robust to noise than other methods. A p-value shows how significant a method on the left column outperforms a method in the top row in terms of cis and trans enrichments. For trans-enrichment, geQTL+ is the best, and MTLasso2G comes in second, outperforming FaSTLMM, SET-eQTL, LORS, Matrix eQTL and Lasso. LORS outperforms Matrix eQTL and Lasso for both cis- and trans-enrichment. This is because LORS considers confounding factors while Matrix eQTL and Lasso does not. These methods each detected about 6000 associations according to non-zero W values. Gene ontology enrichment analysis on detected group-wise associations for yeast We further evaluate the quality of detected groups of genes by measuring the correlations between the detected groups of genes and the GO (Gene Ontology) categories

Background

Methods

VI VII VIIIIX X XI XII XIII XIV XV XVI

Conclusions

X IX VIII VII VI V