Abstract
BackgroundEpistatic interactions of multiple single nucleotide polymorphisms (SNPs) are now believed to affect individual susceptibility to common diseases. The detection of such interactions, however, is a challenging task in large scale association studies. Ant colony optimization (ACO) algorithms have been shown to be useful in detecting epistatic interactions.FindingsAntEpiSeeker, a new two-stage ant colony optimization algorithm, has been developed for detecting epistasis in a case-control design. Based on some practical epistatic models, AntEpiSeeker has performed very well.ConclusionsAntEpiSeeker is a powerful and efficient tool for large-scale association studies and can be downloaded from http://nce.ads.uga.edu/~romdhane/AntEpiSeeker/index.html.
Highlights
Epistatic interactions of multiple single nucleotide polymorphisms (SNPs) are believed to affect individual susceptibility to common diseases
As defined by Dorigio and Gambardella [13], Ant colony optimization (ACO) is comprised of parallel artificial ants that communicate through a probability density function (PDF) that is updated by weights or 'pheromone levels'
We suggest two rounds of search: 1) using a relatively large size SNP set, which is sensitive to strong signals, AntEpiSeeker algorithm Input paramters D: a dataset of N case and control samples genotyped at L loci iEpiModel: number of SNPs in an epistatic interaction Pvalue: statistical significance threshold largesetsize: size of the large SNP sets smallsetsize: size of the small SNP sets iAntCount: number of ants iItCountLarge, iItCountSmall: number of iterations for large or small SNP sets τ 0 : initial pheromone level for each locus ρ : evaporation rate α : parameter determining the weight given to pheromone deposited by ants for setsize in If setsize==largesetsize, iItCount=iItCountLarge; else, iItCount=iItCountSmall
Summary
Genetic association studies, which aim at detecting association between one or more genetic polymorphisms and a trait of interest such as a quantitative characteristic, discrete attribute or disease, have gained a lot of popularity in the past decade [1]. An exhaustive search of two-locus interactions needs to evaluate at least 5.00 × 109 locus combinations, and this number increases to 1.67 × 1014 when three-locus interactions are considered This process is computationally hard it could be enhanced by two recent approaches: the Bayesian epistasis association mapping (BEAM) [11] and SNPharvester [12], which were shown to be able to handle large scale datasets. The use of MDR for detecting epistatic interactions in these studies dramatically increased the computational burden These studies did not test performance using the more practical epistatic models such as the ones proposed by Marchini et al [16]. The twostage design of ant colony optimization and the idea of searching bigger SNP sets harboring epistatic interactions enhance the power of ACO algorithms. AntEpiSeeker showed improved performance based on some practical epistatic models and large scale datasets
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call