Email updates

Keep up to date with the latest news and content from BioData Mining and BioMed Central.

Open Access Methodology

Application of a spatially-weighted Relief algorithm for ranking genetic predictors of disease

Matthew E Stokes* and Shyam Visweswaran

Author Affiliations

Department of Biomedical Informatics and the Intelligent Systems Program, University of Pittsburgh, 5607 Baum Boulevard, Pittsburgh, PA, 15206, USA

For all author emails, please log on.

BioData Mining 2012, 5:20  doi:10.1186/1756-0381-5-20

Published: 3 December 2012

Abstract

Background

Identification of genetic variants that are associated with disease is an important goal in elucidating the genetic causes of diseases. The genetic patterns that are associated with common diseases are complex and may involve multiple interacting genetic variants. The Relief family of algorithms is a powerful tool for efficiently identifying genetic variants that are associated with disease, even if the variants have nonlinear interactions without significant main effects. Many variations of Relief have been developed over the past two decades and several of them have been applied to single nucleotide polymorphism (SNP) data.

Results

We developed a new spatially weighted variation of Relief called Sigmoid Weighted ReliefF Star (SWRF*), and applied it to synthetic SNP data. When compared to ReliefF and SURF*, which are two algorithms that have been applied to SNP data for identifying interactions, SWRF* had significantly greater power. Furthermore, we developed a framework called the Modular Relief Framework (MoRF) that can be used to develop novel variations of the Relief algorithm, and we used MoRF to develop the SWRF* algorithm.

Conclusions

MoRF allows easy development of new Relief algorithms by specifying different interchangeable functions for the component terms. Using MORF, we developed a new Relief algorithm called SWRF* that had greater ability to identify interacting genetic variants in synthetic data compared to existing Relief algorithms.

Keywords:
Feature selection; Relief; Single nucleotide polymorphisms; Genetic interactions; Epistasis