Investigating statistical epistasis in complex disorders

James C. Turton, James Bullock, Christopher Medway, Hui Shi, Kristelle Brown, Olivia Belbin, Noor Kalsheker, Minerva M. Carrasquillo, Dennis W. Dickson, Neill R. Graff-Radford, Ronald C. Petersen, Steven G. Younkin, Kevin Morgan

Research output: Contribution to journalArticle

4 Scopus citations

Abstract

The missing heritability exhibited by late-onset Alzheimer's disease is unexplained and has been partly attributed to epistatic interaction. Methods available to explore this are often based on logistic regression and allow for determination of deviation from an expected outcome as a result of statistical epistasis. Three such methodologies including Synergy Factor and the PLINK modules,-epistasis and-fast-epistasis, were applied to study an epistatic interaction between interleukin-6 and interleukin-10. The models analyzed consisted of two synergistic interactions (SF ≈ 4.2 and 1.6) and two antagonistic interactions (SF ≈ 0.9 and 0.6). As with any statistical test, power to detect association is paramount; and most studies will be underpowered for the task. However, the availability of large sample sizes through genome-wide association studies make it feasible to examine approaches for determining epistatic interactions. This study documents the sample sizes needed to achieve a statistically significant outcome from each of the methods examined and discusses the limitations/advantages of the chosen approaches.

Original languageEnglish (US)
Pages (from-to)635-644
Number of pages10
JournalJournal of Alzheimer's Disease
Volume25
Issue number4
DOIs
StatePublished - 2011

Keywords

  • Complex disorders
  • LOAD
  • PLINK
  • epistasis
  • modeling
  • synergy factor

ASJC Scopus subject areas

  • Neuroscience(all)
  • Clinical Psychology
  • Geriatrics and Gerontology
  • Psychiatry and Mental health

Fingerprint Dive into the research topics of 'Investigating statistical epistasis in complex disorders'. Together they form a unique fingerprint.

  • Cite this