LLR: A latent low-rank approach to colocalizing genetic risk variants in multiple GWAS

Jin Liu, Xiang WAN, Chaolong Wang, Chao Yang, Xiaowei Zhou, Can YANG*

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

3 Citations (Scopus)

Abstract

Motivation: Genome-wide association studies (GWAS), which genotype millions of single nucleotide polymorphisms (SNPs) in thousands of individuals, are widely used to identify the risk SNPs underlying complex human phenotypes (quantitative traits or diseases). Most conventional statistical methods in GWAS only investigate one phenotype at a time. However, an increasing number of reports suggest the ubiquity of pleiotropy, i.e. many complex phenotypes sharing common genetic bases. This motivated us to leverage pleiotropy to develop new statistical approaches to joint analysis of multiple GWAS. Results: In this study, we propose a latent low-rank (LLR) approach to colocalizing genetic risk variants using summary statistics. In the presence of pleiotropy, there exist risk loci that affect multiple phenotypes. To leverage pleiotropy, we introduce a low-rank structure to modulate the probabilities of the latent association statuses between loci and phenotypes. Regarding the computational efficiency of LLR, a novel expectation-maximization-path (EM-path) algorithm has been developed to greatly reduce the computational cost and facilitate model selection and inference. We demonstrate the advantages of LLR over competing approaches through simulation studies and joint analysis of 18 GWAS datasets.

Original languageEnglish
Pages (from-to)3878-3886
Number of pages9
JournalBioinformatics
Volume33
Issue number24
DOIs
Publication statusPublished - 15 Dec 2017

Scopus Subject Areas

  • Statistics and Probability
  • Biochemistry
  • Molecular Biology
  • Computer Science Applications
  • Computational Theory and Mathematics
  • Computational Mathematics

Fingerprint

Dive into the research topics of 'LLR: A latent low-rank approach to colocalizing genetic risk variants in multiple GWAS'. Together they form a unique fingerprint.

Cite this