ATHENA: A knowledge-based hybrid backpropagation-grammatical evolution neural network algorithm for discovering epistasis among quantitative trait Loci.

Stephen D Turner,Scott M Dudek,Marylyn D Ritchie

doi:10.1186/1756-0381-3-5

Stephen D Turner, Scott M Dudek + Show 1 more

Open Access

https://doi.org/10.1186/1756-0381-3-5

Copy DOI

Journal: BioData mining	Publication Date: Sep 27, 2010
Citations: 111	License type: cc-by

Affiliation: Vanderbilt University, Center for Human Genetics

Abstract

BackgroundGrowing interest and burgeoning technology for discovering genetic mechanisms that influence disease processes have ushered in a flood of genetic association studies over the last decade, yet little heritability in highly studied complex traits has been explained by genetic variation. Non-additive gene-gene interactions, which are not often explored, are thought to be one source of this "missing" heritability.MethodsStochastic methods employing evolutionary algorithms have demonstrated promise in being able to detect and model gene-gene and gene-environment interactions that influence human traits. Here we demonstrate modifications to a neural network algorithm in ATHENA (the Analysis Tool for Heritable and Environmental Network Associations) resulting in clear performance improvements for discovering gene-gene interactions that influence human traits. We employed an alternative tree-based crossover, backpropagation for locally fitting neural network weights, and incorporation of domain knowledge obtainable from publicly accessible biological databases for initializing the search for gene-gene interactions. We tested these modifications in silico using simulated datasets.ResultsWe show that the alternative tree-based crossover modification resulted in a modest increase in the sensitivity of the ATHENA algorithm for discovering gene-gene interactions. The performance increase was highly statistically significant when backpropagation was used to locally fit NN weights. We also demonstrate that using domain knowledge to initialize the search for gene-gene interactions results in a large performance increase, especially when the search space is larger than the search coverage.ConclusionsWe show that a hybrid optimization procedure, alternative crossover strategies, and incorporation of domain knowledge from publicly available biological databases can result in marked increases in sensitivity and performance of the ATHENA algorithm for detecting and modelling gene-gene interactions that influence a complex human trait.

Highlights

Growing interest and burgeoning technology for discovering genetic mechanisms that influence disease processes have ushered in a flood of genetic association studies over the last decade, yet little heritability in highly studied complex traits has been explained by genetic variation
Genome-Wide Association Studies, Complex Disease, and Epistasis The genome-wide association study (GWAS) is a widely used technique in human genetics research to investigate DNA variations associated with common human diseases
The last several decades have ushered in technological advances that have allowed investigators to progress from coarse genomic coverage with linkage maps and candidate gene association studies, to very high resolution association studies using single nucleotide polymorphisms (SNPs)

Summary

Introduction

Growing interest and burgeoning technology for discovering genetic mechanisms that influence disease processes have ushered in a flood of genetic association studies over the last decade, yet little heritability in highly studied complex traits has been explained by genetic variation. The most commonly used analytical procedures for analyzing GWAS data are very simple tests of association looking at one SNP at a time This approach has been somewhat successful in identifying genetic variants associated with complex traits, including age-related macular degeneration [4], type II diabetes [5], hypertension [6], and blood cholesterol levels [7,8], among others [9]. These single SNPs collectively explain little of the genetic contribution to the trait variance that is expected based on family and twin studies [10]. Several recent perspectives have emphasized that most true single locus genetic associations to complex traits carry a vanishingly small effect size [18,19], and experimental data from model organisms illustrates that gene-gene interaction is pervasive and often carries surprisingly large effects [20,21]

Methods

Results

Conclusion