R Code for MT-eQTL

An Empirical Bayes Approach for Multiple Tissue eQTL Analysis

Gen Li external link, Andrey A. Shabalin external link, Ivan Rusyn external link, Fred A. Wright external link, Andrew B. Nobel external link

Applied to the data from the GTEx project external link.

Based on Matrix eQTL.

The manuscript is available at arxiv.org external link

Features of MT-eQTL model

  • Multivariate Bayesian Hierarchical Model.
  • Model captures presence or absence of eQTLs in each tissue, and heterogeneity of effect sizes.
  • Number and identity of donors can vary from tissue to tissue.
  • Enhances statistical power for single and multiple tissue analyses.

Features of MT-eQTL code

  • Accepts standard output from Matrix eQTL.
  • Memory efficient.
  • EM algorithm for fast maximum likelihood estimation.
Performance:
Number of tissues time
1 15 minutes
9 ~24 hours

Data: GTEx genotype and expression data with 10,559,127 local gene-SNP pairs across 9 tissues.

Specifications: Intel i7-3770 CPU, 8GB RAM, Windows 7, Revolution R 6.2. 		Steps of the analysis:

R logo icon  1. Run Matrix eQTL for each tissue

R logo icon  2. Combine results across tissues in one matrix

R logo icon  3. Estimate the model parameters

R logo icon  4. Call eQTLs

help icon  Frequently asked questions

Questions, comments, concerns?
Contact me: Andrey A. Shabalin



By Andrey Shabalin