Learn R Programming

RFmarkerDetector (version 1.0.1)

Multivariate Analysis of Metabolomics Data using Random Forests

Description

A collection of tools for multivariate analysis of metabolomics data, which includes several preprocessing methods (normalization, scaling) and various exploration and data visualization techniques (Principal Components Analysis and Multi Dimensional Scaling). The core of the package is the Random Forest algorithm used for the construction, optimization and validation of classification models with the aim of identifying potentially relevant biomarkers.

Copy Link

Version

Install

install.packages('RFmarkerDetector')

Monthly Downloads

21

Version

1.0.1

License

GPL-3

Maintainer

Piergiorgio Palla

Last Published

February 29th, 2016

Functions in RFmarkerDetector (1.0.1)

AUC multiple cross-validation

forestPerformance

Characterizing the performance of a Random Forest model

Unit variance scaling method performed on the columns of the data (i.e. metabolite concentrations measured by 1H NMR or binned 1H NMR spectra)

Metabolite concentrations

combinatorialRFMCCV

Combinatorial Monte Carlo CV

Filtering 'low quality' variables from the original dataset

Computing the average AUC

Extracting the best performing Random Forest model

Mtry Optimization

plotAUCvsCombinations

Plotting the average AUC as a function of the number of combinations

Variable Frequency Plot

Mean centering performed on the columns of the data (i.e. metabolite concentrations measured by 1H NMR or binned 1H NMR spectra)

plot.pca.scores

PCA Scores plot This function creates a plot that graphically projects the original samples onto the subspce spanned by the first two principal components

Pareto scaling method performed on the columns of the data table (i.e. metabolite concentrations measured by 1H NMR or binned 1H NMR spectra)

Tuning of the ntree parameter (i.e. the number of trees) for a Random Forest model

Multi-dimensional Scaling (MDS) Plot

plot.pca.loadings

PCA Loadings plot This function plots the relation between the original variables and the subspace dimensions. It is useful for interpreting relationships among variables.

Plotting single or multiple ROC curves of the cross-validated Random Forest models plot.mccv allows to plot single or multiple ROC curves to characterize the performace of a cross-validated Random Forest model

Principal Component Analysis

Plotting the average OOB error and its 95% confidence interval as a function of the mtry parameter

Extracting average accuracy and recall of a list of Random Forest models

Filtering less informative variables

Computing relative standard deviation of a vector

Tuning of the mtry parameter for a Random Forest model

Monte Carlo cross-validation of Random Forest models