The goal of this article (letter to the editor) is to emphasize the value of exploring ranking stability when using the importance measures, mean decrease accuracy (MDA) and mean decrease Gini (MDG), provided by Random Forest. We illustrate with a real and a simulated example that ranks based on the MDA are unstable to small perturbations of the dataset and ranks based on the MDG provide more robust results.

Objective: Genomic profiling, the use of genetic variants at multiple loci simultaneously for the prediction of disease risk, requires the selection of a set of genetic variants that best predicts disease status. The goal of this work was to provide a new selection algorithm for genomic profiling. Methods: We propose a new algorithm for genomic profiling based on optimizing the area under the receiver operating characteristic curve (AUC) of the random forest (RF). The proposed strategy implements a backward elimination process based on the initial ranking of variables. Results and Conclusions: We demonstrate the advantage of using the AUC instead of the classification error as a measure of predictive accuracy of RF. In particular, we show that the use of the classification error is especially inappropriate when dealing with unbalanced data sets. The new procedure for variable selection and prediction, namely AUC-RF, is illustrated with data from a bladder cancer study and also with simulated data. The algorithm is publicly available as an R package, named AUCRF, at http://cran.r-project.org/.

In rhesus macaques, simian immunodeficiency virus infection is followed by expansion of enteric viruses but has a limited impact on the gut bacteriome. To understand the longitudinal effects of HIV-1 infection on the human gut microbiota, we prospectively followed 49 Mozambican subjects diagnosed with recent HIV-1 infection (RHI) and 54 HIV-1-negative controls for 9–18 months and compared them with 98 chronically HIV-1-infected subjects treated with antiretrovirals (n = 27) or not (n = 71). We show that RHI is followed by increased fecal adenovirus shedding, which persists during chronic HIV-1 infection and does not resolve with ART. Recent HIV-1 infection is also followed by transient non-HIV-specific changes in the gut bacterial richness and composition. Despite early resilience to change, an HIV-1-specific signature in the gut bacteriome—featuring depletion of Akkermansia, Anaerovibrio, Bifidobacterium, and Clostridium—previously associated with chronic inflammation, CD8+ T cell anergy, and metabolic disorders, can be eventually identified in chronically HIV-1-infected subjects. Recent HIV-1 infection is associated with increased fecal shedding of eukaryotic viruses, transient loss of bacterial taxonomic richness, and long-term reductions in microbial gene richness. An HIV-1-associated microbiome signature only becomes evident in chronically HIV-1-infected subjects.

Summary: We describe mbmdr, an R package for implementing the model-based multifactor dimensionality reduction (MB-MDR) method. MB-MDR has been proposed by Calle et al. as a dimension reduction method for exploring gene–gene interactions in case-control association studies. It is an extension of the popular multifactor dimensionality reduction (MDR) method of Ritchie et al. allowing a more flexible definition of risk cells. In MB-MDR, risk categories are defined using a regression model which allows adjustment for covariates and main effects and, in addition to the classical low risk and high risk categories, MB-MDR considers a third category of indeterminate or not informative cells. An important improvement added to the current mbmdr algorithm with respect to the original MB-MDR formulation in Calle et al. and also to the classical MDR approach, is the extension of the methodology to different outcome types. While MB-MDR was initially proposed for binary traits in the context of case-control studies, the mbmdr package provides options to analyze both binary or quantitative traits for unrelated individuals. Availability:http://cran.r-project.org/ Contact:malu.calle@uvic.cat Supplementary information: Supplementary data are available at Bioinformatics online.