PanelomiX

The R package to create panels of biomarkers with PanelomiX.

For more information, see:

1. Xavier Robin, Natacha Turck, Alexandre Hainard, et al. (2013) “PanelomiX: A threshold-based algorithm to create panels of biomarkers”. Translational Proteomics, 1 (1), 57-64. DOI: 10.1016/j.trprot.2013.04.003
2. The official web page

Experimental

This package is highly experimental. For an easier experience please use the web-based tool. Functions and arguments may change at any time and without notice. Functionality is not guaranteed. Feel free to report bugs on the bug tracker. However this project is no longer under active development, and support and bug fixes are not guaranteed.

To install this package:

    if (! requireNamespace("devtools")) install.packages("devtools")
    devtools::install_github("xrobin/PanelomiX")

System Requirements

This package requires a GNU/Linux system with Java installed and a working installation of rJava.

Windows is supported. MacOS is untested.

Help

Once the library is loaded with library(PanelomiX), you can get help on PanelomiX by typing ?PanelomiX.

Getting started

If you don't want to read the manual first, try the following:

Loading some data

library(PanelomiX)
library(pROC)
data(aSAH)

Basic training of a panel

panels <- exh.train( data = aSAH, 
                     predictors = c("age", "s100b", "ndka"), 
                     response = "outcome")

Test a specific threshold for one/many candidate predictor(s)

panels <- exh.train(data = aSAH, 
                    predictors = c("age", "s100b", "ndka"), 
                    response = "outcome", 
                    test.thresholds.predictors = list(s100b = 0.5))

Test a specific direction for one/many candidate predictor(s)

panels <- exh.train(data = aSAH, 
                    predictors = c("age", "s100b", "ndka"), 
                    fixed.predictors = c("age"), 
                    response = "outcome",
                    directions = list(age = ">"))

Cross-validation

cv <- exh.train.cv(aSAH, c("age", "s100b", "ndka"), "outcome", progress=FALSE)

Predict new data

predict(panels, aSAH)

Count how many panels were found

nr.panels(panels)
nr.panels(cv)

Inspect the stability of the cross-validation

# Biomarker selection
table.mol.stability(cv)
# Number of biomarkers selected in the panel
table.nr.stability(cv)
# How many biomarkers should be positive for the panel to be positive?
table.min.nr.stability(cv)

Development

To rebuild the java package:

cd java
make 

Then you can install the package normally.