Functional regression method for whole genome eQTL epistasis analysis with sequencing data

BMC Genomics

Methodology Article

Li Jin¹  Momiao Xiong²  Kelin Xu³ 

[1] State Key Laboratory of Genetic Engineering and Ministry of Education Key Laboratory of Contemporary Anthropology, Collaborative Innovation Center for Genetics and Development, School of Life Sciences, Fudan University, 200438, Shanghai, China;State Key Laboratory of Genetic Engineering and Ministry of Education Key Laboratory of Contemporary Anthropology, Collaborative Innovation Center for Genetics and Development, School of Life Sciences, Fudan University, 200438, Shanghai, China;Department of Biostatistics, Human Genetics Center, The University of Texas Health Science Center at Houston, 77030, Houston, TX, USA;Human Genetics Center, The University of Texas Health Science Center at Houston, P.O. Box 20186, 77225, Houston, TX, USA;State Key Laboratory of Genetic Engineering and Ministry of Education Key Laboratory of Contemporary Anthropology, Collaborative Innovation Center for Genetics and Development, School of Life Sciences, Fudan University, 200438, Shanghai, China;School of Data Science and Institute for Big Data, Fudan University, 200433, Shanghai, China;

关键词: Gene-gene interaction;    Multivariate functional regression;    Functional regression models;    RNA-seq;    Next-generation sequencing;    Association studies;    eQTL;   

DOI  :  10.1186/s12864-017-3777-4

 received in 2016-11-21, accepted in 2017-05-09,  发布年份 2017

来源: Springer

【 摘 要 】

BackgroundEpistasis plays an essential rule in understanding the regulation mechanisms and is an essential component of the genetic architecture of the gene expressions. However, interaction analysis of gene expressions remains fundamentally unexplored due to great computational challenges and data availability. Due to variation in splicing, transcription start sites, polyadenylation sites, post-transcriptional RNA editing across the entire gene, and transcription rates of the cells, RNA-seq measurements generate large expression variability and collectively create the observed position level read count curves. A single number for measuring gene expression which is widely used for microarray measured gene expression analysis is highly unlikely to sufficiently account for large expression variation across the gene. Simultaneously analyzing epistatic architecture using the RNA-seq and whole genome sequencing (WGS) data poses enormous challenges.MethodsWe develop a nonlinear functional regression model (FRGM) with functional responses where the position-level read counts within a gene are taken as a function of genomic position, and functional predictors where genotype profiles are viewed as a function of genomic position, for epistasis analysis with RNA-seq data. Instead of testing the interaction of all possible pair-wises SNPs, the FRGM takes a gene as a basic unit for epistasis analysis, which tests for the interaction of all possible pairs of genes and use all the information that can be accessed to collectively test interaction between all possible pairs of SNPs within two genome regions.ResultsBy large-scale simulations, we demonstrate that the proposed FRGM for epistasis analysis can achieve the correct type 1 error and has higher power to detect the interactions between genes than the existing methods. The proposed methods are applied to the RNA-seq and WGS data from the 1000 Genome Project. The numbers of pairs of significantly interacting genes after Bonferroni correction identified using FRGM, RPKM and DESeq were 16,2361, 260 and 51, respectively, from the 350 European samples.ConclusionsThe proposed FRGM for epistasis analysis of RNA-seq can capture isoform and position-level information and will have a broad application. Both simulations and real data analysis highlight the potential for the FRGM to be a good choice of the epistatic analysis with sequencing data.

【 授权许可】

CC BY   
© The Author(s). 2017

【 图 表 】

Fig. 4

Fig. 10

Fig. 1

Fig. 1

Fig. 2

Scheme 1

12951_2016_223_Article_IEq1.gif

12951_2017_255_Article_IEq50.gif

12951_2017_255_Article_IEq49.gif

12936_2017_1904_Article_IEq1.gif

Fig. 1

Fig. 5

42004_2023_1031_Figa_HTML.png

12951_2016_246_Article_IEq9.gif

42004_2023_1031_Article_IEq16.gif

12951_2016_246_Article_IEq8.gif

Fig. 5

Fig. 1

Fig. 2

Fig. 10