Gene - Gene Interactions Detection Using a Two-stage Model

Zhanyong Wang, Jae Hoon Sul, Sagi Snir, Jose A. Lozano, Eleazar Eskin

Research output: Contribution to journalArticlepeer-review


Genome-wide association studies (GWAS) have discovered numerous loci involved in genetic traits. Virtually all studies have reported associations between individual single nucleotide polymorphisms (SNPs) and traits. However, it is likely that complex traits are influenced by interaction of multiple SNPs. One approach to detect interactions of SNPs is the brute force approach which performs a pairwise association test between a trait and each pair of SNPs. The brute force approach is often computationally infeasible because of the large number of SNPs collected in current GWAS studies. We propose a two-stage model, Threshold-based Efficient Pairwise Association Approach (TEPAA), to reduce the number of tests needed while maintaining almost identical power to the brute force approach. In the first stage, our method performs the single marker test on all SNPs and selects a subset of SNPs that achieve a certain significance threshold. In the second stage, we perform a pairwise association test between traits and pairs of the SNPs selected from the first stage. The key insight of our approach is that we derive the joint distribution between the association statistics of a single SNP and the association statistics of pairs of SNPs. This joint distribution allows us to provide guarantees that the statistical power of our approach will closely approximate the brute force approach. We applied our approach to the Northern Finland Birth Cohort data and achieved 63 times speedup while maintaining 99% of the power of the brute force approach.

Original languageEnglish
Pages (from-to)563-576
Number of pages14
JournalJournal of Computational Biology
Issue number6
StatePublished - 1 Jun 2015

Bibliographical note

Publisher Copyright:
© Copyright 2015, Mary Ann Liebert, Inc.


  • GWAS
  • epistasis
  • gene-gene interaction

ASJC Scopus subject areas

  • Computational Mathematics
  • Genetics
  • Molecular Biology
  • Computational Theory and Mathematics
  • Modeling and Simulation


Dive into the research topics of 'Gene - Gene Interactions Detection Using a Two-stage Model'. Together they form a unique fingerprint.
  • Gene-gene interactions detection using a two-stage model

    Wang, Z., Sul, J. H., Snir, S., Lozano, J. A. & Eskin, E., 2014, Research in Computational Molecular Biology - 18th Annual International Conference, RECOMB 2014, Proceedings. Springer Verlag, p. 340-355 16 p. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); vol. 8394 LNBI).

    Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Cite this