Identifying SNPs predictive of phenotype using random forests

Abstract

There has been a great interest and a few successes in the identification of complex disease susceptibility genes in recent years. Association studies, where a large number of single-nucleotide polymorphisms (SNPs) are typed in a sample of cases and controls to determine which genes are associated with a specific disease, provide a powerful approach for complex disease gene mapping. Genes of interest in those studies may contain large numbers of SNPs that classical statistical methods cannot handle simultaneously without requiring prohibitively large sample sizes. By contrast, high-dimensional nonparametric methods thrive on large numbers of predictors. This work explores the application of one such method, random forests, to the problem of identifying SNPs predictive of the phenotype in the case-control study design. A random forest is a collection of classification trees grown on bootstrap samples of observations, using a random subset of predictors to define the best split at each node. The observations left out of the bootstrap samples are used to estimate prediction error. The importance of a predictor is quantified by the increase in misclassification occurring when the values of the predictor are randomly permuted. We extend the concept of importance to pairs of predictors, to capture joint effects, and we explore the behavior of importance measures over a range of two-locus disease models in the presence of a varying number of SNPs unassociated with the phenotype. We illustrate the application of random forests with a data set of asthma cases and unaffected controls genotyped at 42 SNPs in ADAM33, a previously identified asthma susceptibility gene. SNPs and SNP pairs highly associated with asthma tend to have the highest importance index value, but predictive importance and association do not always coincide.

Patterns of GATA3 and IL13 gene polymorphisms associated with childhood rhinitis and atopy in a birth cohort

“…31 There has been great interest in the random forest classification procedure in studies with a large number of SNPs. 34 In unbalanced association studies the misclassification rate can be high, and therefore we use the random forest ranking only in combination with other criteria. Use of a 2-step procedure, with model selection based on AIC and a ranking function for the SNPs for validation, reduces the effect of selecting a model that by chance produces the largest difference between cases and control subjects.…”

Random forest fishing: a novel approach to identifying organic group of risk factors in genome-wide association studies

“…Use of the importance measure precludes the need to explicitly model every possible interaction terms; and makes interaction analysis of many variables less strenuous. RF was shown to perform well by simulation, 13 and in genetic studies with moderate number of variables, including microarray data analysis 14,15 and association analyses with no more than hundreds of SNPs. 13,16,17 It can effectively select the few important variables out from a large number of irrelevant ones (noise), and be used when the number of variables is much larger than the number of observations.…”

“…RF was shown to perform well by simulation, 13 and in genetic studies with moderate number of variables, including microarray data analysis 14,15 and association analyses with no more than hundreds of SNPs. 13,16,17 It can effectively select the few important variables out from a large number of irrelevant ones (noise), and be used when the number of variables is much larger than the number of observations. Recent advances such as Random Jungle (RJ) 18 have made it possible to construct large RFs from genome-wide data.…”

Pathway-based identification of SNPs predictive of survival

“…[9][10][11] Overall, random forests is among the best approaches for analyzing survival time using gene expression data. [12][13][14] In this article, we introduce one of the first methods to correlate SNP with survival outcomes.…”