sourcesinc Files

Predicting novel microRNAs: a comprehensive comparison of machine learning approaches

• G. Stegmayer, L. Di Persia, M. Rubiolo, M. Gerard, C. Yones, M. Pividori, L. Bugnon, T. Rodriguez, J. Raad and D.H. Milone
• sinc(i) - http://fich.unl.edu.ar/sinc/
• gstegmayer@sinc.unl.edu.ar

This work is a comprehensive review and comparative assessment of methods from two machine learning paradigms for dealing with the prediction of novel pre-miRNAs: supervised and unsupervised training. They have been compared in several prediction tasks involving two model genomes and several increasing imbalance levels.

Setup and run

Requirements: - Matlab R2014b

Simple test: - Run the main.m script in the Matlab console

Steps to run the full code: - Open the traintest.m file and select (uncomment) one or more of the machine learning algorithms reviewed in this study - Run the script main.m in a Matlab command line in order to reproduce the complete set of results for the method chosen, at all imbalance levels studied and both data sets

Notes

• The source code for methods reviewed in the study are in the methods/ folder
• The data used in this study is in the data/ folder
• For each run of main.m, several log files with the independent cross-validation results and corresponding mean values of the performance measures are stored in the log/ folder
• By default, a 10-fold cross-validation is performed for each method at each imbalance level

WARNING

• Please keep in mind that the whole procedure (training ALL methods and ALL imbalance levels in BOTH datasets) can take several hours/days.