Evaluating the Ability of Tree‐Based Methods and Logistic Regression for the Detection of SNP‐SNP Interaction

Abstract

SummaryMost common human diseases are likely to have complex etiologies. Methods of analysis that allow for the phenomenon of epistasis are of growing interest in the genetic dissection of complex diseases. By allowing for epistatic interactions between potential disease loci, we may succeed in identifying genetic variants that might otherwise have remained undetected. Here we aimed to analyze the ability of logistic regression (LR) and two tree-based supervised learning methods, classification and regression trees (CART) and random forest (RF), to detect epistasis. Multifactor-dimensionality reduction (MDR) was also used for comparison. Our approach involves first the simulation of datasets of autosomal biallelic unphased and unlinked single nucleotide polymorphisms (SNPs), each containing a two-loci interaction (causal SNPs) and 98 'noise' SNPs. We modelled interactions under different scenarios of sample size, missing data, minor allele frequencies (MAF) and several penetrance models: three involving both (indistinguishable) marginal effects and interaction, and two simulating pure interaction effects. In total, we have simulated 99 different scenarios. Although CART, RF, and LR yield similar results in terms of detection of true association, CART and RF perform better than LR with respect to classification error. MAF, penetrance model, and sample size are greater determining factors than percentage of missing data in the ability of the different techniques to detect true association. In pure interaction models, only RF detects association. In conclusion, tree-based methods and LR are important statistical tools for the detection of unknown interactions among true risk-associated SNPs with marginal effects and in the presence of a significant number of noise SNPs. In pure interaction models, RF performs reasonably well in the presence of large sample sizes and low percentages of missing data. However, when the study design is suboptimal (unfavourable to detect interaction in terms of e.g. sample size and MAF) there is a high chance of detecting false, spurious associations.

Overview of random forest methodology and practical guidance with emphasis on computational biology and bioinformatics

“…This good performance is documented in several independent comparison studies implementing different simulation settings [26,54,73,17]. In these studies, however, standard VIMs (either Gini or permutation) are used to rank the SNPs.…”

Random Forests for Genetic Association Studies

“…Multiple studies have compared RF to other parametric and computational methods, showing the variable importance scores to be more powerful and stable than other current approaches ([García-Margariños et al 2009;Lunetta et al 2004;McKinney et al 2006;Nicodemus et al 2007]. In addition to gene x gene interaction RF has also been used to successfully detect gene x environment interactions [Maenner et al 2009].…”

Power of Data Mining Methods to Detect Genetic Associations and Interactions

“…In comparison, García-Magariños et al [4] found that RF and CART performed as well as logistic regression and MDR when there were SNPs with marginal effects and unknown interactions in the presence of a large number of noise SNPs. In pure interaction models, they found that RF performed as well as MDR especially with large sample sizes.…”

“…Lunetta et al [3] show that RF is more efficient than Fisher's exact test for ranking true disease-associated SNPs. García-Magariños et al [4] compare the top-rated SNP as chosen by RF, classification and regression trees (CART), logistic regression, and multifactor dimensionality reduction (MDR) over numerous settings including several with missing data, whereas Szymczak et al [5] compare RF and ensemble methods to penalized regression and network analyses. Jiang et al [6] use RF Gini variable importance measure to rank SNPs before implementing a forward feature selection algorithm to choose a subset of SNPs and then adopt a hierarchical procedure (unrelated to RF) to determine the statistical significance of the subset.…”