A method combining a random forest-based technique with the modeling of linkage disequilibrium through latent variables, to run multilocus genome-wide association studies

Abstract

BackgroundGenome-wide association studies (GWASs) have been widely used to discover the genetic basis of complex phenotypes. However, standard single-SNP GWASs suffer from lack of power. In particular, they do not directly account for linkage disequilibrium, that is the dependences between SNPs (Single Nucleotide Polymorphisms).ResultsWe present the comparative study of two multilocus GWAS strategies, in the random forest-based framework. The first method, T-Trees, was designed by Botta and collaborators (Botta et al., PLoS ONE 9(4):e93379, 2014). We designed the other method, which is an innovative hybrid method combining T-Trees with the modeling of linkage disequilibrium. Linkage disequilibrium is modeled through a collection of tree-shaped Bayesian networks with latent variables, following our former works (Mourad et al., BMC Bioinformatics 12(1):16, 2011). We compared the two methods, both on simulated and real data. For dominant and additive genetic models, in either of the conditions simulated, the hybrid approach always slightly performs better than T-Trees. We assessed predictive powers through the standard ROC technique on 14 real datasets. For 10 of the 14 datasets analyzed, the already high predicted power observed for T-Trees (0.910-0.946) can still be increased by up to 0.030. We also assessed whether the distributions of SNPs’ scores obtained from T-Trees and the hybrid approach differed. Finally, we thoroughly analyzed the intersections of top 100 SNPs output by any two or the three methods amongst T-Trees, the hybrid approach, and the single-SNP method.ConclusionsThe sophistication of T-Trees through finer linkage disequilibrium modeling is shown beneficial. The distributions of SNPs’ scores generated by T-Trees and the hybrid approach are shown statistically different, which suggests complementary of the methods. In particular, for 12 of the 14 real datasets, the distribution tail of highest SNPs’ scores shows larger values for the hybrid approach. Thus are pinpointed more interesting SNPs than by T-Trees, to be provided as a short list of prioritized SNPs, for a further analysis by biologists. Finally, among the 211 top 100 SNPs jointly detected by the single-SNP method, T-Trees and the hybrid approach over the 14 datasets, we identified 72 and 38 SNPs respectively present in the top25s and top10s for each method.Electronic supplementary materialThe online version of this article (10.1186/s12859-018-2054-0) contains supplementary material, which is available to authorized users.

Machine Learning Methods for Diabetes Prediction

“…Boosting algorithms have been used successfully to solve the case of genome-wide data sets due to linkage disequilibrium. Another comparison study that used genetic data has been done by [13]. The new variant Decision Tree (DT) has been proposed and compared with Random forest using Genome-wide association studies (GWASs) dataset to discover the genetic basis of complex phenotypes.…”

Evaluation of Density-Based Spatial Clustering for Identifying Genomic Loci Associated with Ischemic Stroke in Genome-Wide Data

“…It was a part of the hybrid approach to defining latent variables of a Bayesian network to discover the genetic bases of complex phenotypes. As a result, the GWAS approach utilizing DBSCAN to model LD outperformed the approach based on traditional LD modeling through blocks of contiguous SNPs [23]. This investigation forced us to consider density-based spatial algorithms as promising for the purpose of our research.…”

Hybrid of Restricted and Penalized Maximum Likelihood Method for Efficient Genome-Wide Association Study

“…To address this issue effectively, some improved multi-locus GWAS methods were proposed. For example, multi-locus mixed-model (MLMM) [26] adopts stepwise mixed-model regression with forward inclusion and backward elimination using a Bayesian approach and performs well when the structure is complex, fixed and random model circulating probability unification (FarmCPU) [27] incorporates multiple markers simultaneously as covariates in a stepwise LMM to partially remove the confounding between testing markers and kinship, iterative nonlocal prior-based selection (GWASinlps) [28] considers an iterative structured screen-and-select strategy and nonlocal priors within it and provides an efficient and parsimonious variable selection for continuous phenotypes, a machine-learning method combines a random forest-based technique with the modeling of linkage disequilibrium through latent variables [29] and accelerates the computing speed for multi-locus GWAS, a gene set analysis with Generalized Berk-Jones (GBJ) statistic [30] introduces a permutation-free parametric framework, which can increase the power by incorporating information from multiple signals in the same gene, the SNP set GWAS approach RAINBOW [31] achieves faster computation by using linear kernel for constructing the Gram matrix of the SNP set of interest, and the multi-locus random SNP effect mixed linear model (mrMLM) [32] uses the Wald test based on a random SNP effect linear mixed model to reduce dimensionality; then, all the selected markers are placed into an empirical Bayes [33] multi-locus model, showing the advantage in controlling a complex population structure. A limitation of Bayesian method is that Markov Chain Monte Carlo (MCMC) sampling comes at the cost of intensive computation, or the posterior distribution of fitness is not easy to calculate [34].…”