INFERRING GENE REGULATORY NETWORKS BY MACHINE LEARNING METHODS

The ability to measure the transcriptional response after a stimulus has drawn much attention to the underlying gene regulatory networks. Several machine learning related methods, such as Bayesian networks and decision trees, have been proposed to deal with this difficult problem, but rarely a systematic comparison between different algorithms has been performed. In this work, we critically evaluate the application of multiple linear regression, SVMs, decision trees and Bayesian networks to reconstruct the budding yeast cell cycle network. The performance of these methods is assessed by comparing the topology of the reconstructed models to a validation network. This validation network is defined a priori and each interaction is specified by at least one publication. We also investigate the quality of the network reconstruction if a varying amount of gene regulatory dependencies is provided a priori.