A Review for Detecting Gene-Gene Interactions Using Machine Learning Methods in Genetic Epidemiology

Table 1

Summary of detect gene-gene interaction using neural network method.

No.	Author	Dataset	Description
(1)	Ritchie et al. [11]	Epistasis model.	GPNN and BPNN were used to model gene-gene interactions by using simulated data. The simulated data contains functional SNPs and nonfunctional SNPs which model the interaction between genes.
(2)	Tomita et al. [12]	Childhood allergic asthma (CAA).	Artificial neural network was utilized with parameter decreasing method in order to analyse susceptible SNPs among the Japanese people.
(3)	Keedwell and Narayanan [13]	Artificial data experiments, rat spinal cord and yeast Saccharomyces Cerevisiae cell cycle.	Genetic algorithm which was implemented along with neural networks discovers gene-gene interactions in temporal gene expression dataset by elucidating the information between regulatory connections and interactions between genes, proteins, and other gene products.
(4)	Motsinger et al. [14]	Parkinson’s disease.	GPNN had been used to optimize the architecture of neural network. This method can be used to enhance the identification of gene combinations associated with Parkinson’s disease.
(5)	Ritchie et al. [15]	Alzheimer’s disease, breast’s disease, colorectal disease, and prostate’s disease.	GPNN had been used to detect gene-gene interactions and gene-environment interaction in studies of human disease to optimize the architecture of Neural Network by using simulated dataset.
(6)	Motsinger-Reif et al. [16]	Epitasis model.	GENN was utilized to discover gene-gene interactions that caused are by noise (for instance, genotyping error, missing data, phenocopy, and genetic heterogeneity) in high dimensional genetic epidemiological data.
(7)	Günther et al. [17]	Two-locus disease models, multiplicative and epistasis model.	NN had been used in simulation study to model the different kind of two-locus disease model by constructing six neural networks.
(8)	Turner et al. [18]	Simulated human.	ATHENA had been used to discover the gene-gene interactions that influence complex human traits by integrating alternative tree-based crossover, back propagation, and domain knowledge in ATHENA.
(9)	Hardison and Motsinger-Reif [4]	Genetic models.	QTGENN had applied GENN methods to quantitative traits in various types of simulated genetic models. This method had been successfully applied in single-locus models and two-locus models.