Feature selection with the R package MXM

Abstract

Feature (or variable) selection is the process of identifying the minimal set of features with the highest predictive performance on the target variable of interest. Numerous feature selection algorithms have been developed over the years, but only few have been implemented in R and made publicly available R as packages while offering few options. The R package MXM offers a variety of feature selection algorithms, and has unique features that make it advantageous over its competitors: a) it contains feature selection algorithms that can treat numerous types of target variables, including continuous, percentages, time to event (survival), binary, nominal, ordinal, clustered, counts, left censored, etc; b) it contains a variety of regression models that can be plugged into the feature selection algorithms (for example with time to event data the user can choose among Cox, Weibull, log logistic or exponential regression); c) it includes an algorithm for detecting multiple solutions (many sets of statistically equivalent features, plain speaking, two features can carry statistically equivalent information when substituting one with the other does not effect the inference or the conclusions); and d) it includes memory efficient algorithms for high volume data, data that cannot be loaded into R (In a 16GB RAM terminal for example, R cannot directly load data of 16GB size. By utilizing the proper package, we load the data and then perform feature selection.). In this paper, we qualitatively compare MXM with other relevant feature selection packages and discuss its advantages and disadvantages. Further, we provide a demonstration of MXM’s algorithms using real high-dimensional data from various applications.

Marker genes of incident type 1 diabetes in peripheral blood mononuclear cells of children: A machine learning strategy for large-p, small-n scenarios

“…• Max-Min Parents and Children (MMPC): Incorporated in the MXM R-package [15], this algorithm conducts a constraint-based feature selection, assuming a Bayesian network for input variables. The permutation option was activated (R = 999) and the max_k value was specified as: sample size/10 for optimizing the performance.…”

“…Being counterfactual to the common phenomenon of large-n, smallp contexts, classical statistical models are challenged by this unique structure. Various techniques proposed to address the curse of dimensionality in omics data, are dominated by feature selection workflows [12] and regularization techniques [13], while de-novo algorithms tailored to omics data have also been developed [14,15]. Although a single best method to deal with high-dimensionality does not exist, proposed approaches have unequivocally contributed to increasing the robustness of omics data analytics and biomarker discovery.…”

Current Achievements and Applications of Transcriptomics in Personalized Cancer Medicine

“…Therefore, in current biomarker research simple solutions such as NTP classifier became popular [ 125 , 126 ]. Moreover, sets of selected markers offer much better performance than single marker, therefore, currently in various studies predictive power have benefited from such multi-marker approach [ 127 ]. For example, in the study of Xing et al, three serum markers were evaluated as potentially clinically usable: CHI3L1, MMP13, and SPP1, from more than 4000 differentially expressed genes in the ESCC transcriptome database.…”

Identification of the Prognosis-Related lncRNAs and Genes in Gastric Cancer

“…The predicted risk score for the patients was calculated based on the expression status of the prognostic genes, implemented in R with the predict.coxph function. Particularly, the features (prognostic genes) were selected by the Maximum Minimum Parents and Children (MMPC) algorithm (Lagani et al, 2016) and implemented by the MXM package in R.…”