Experimental Comparison of Symbolic Learning Programs for the Classification of Gene Network Topology Models

This study addresses the problem of identifying the large-scale topology of gene regulation networks from features that can be derived from microarray data sets. Understanding large-scale structures of gene regulation is fundamentally important in biology. Recent analysis of network properties of known biological networks has shown that they display scale-free features, but it is not yet clear whether these features are generic to all biological networks. In this work five different symbolic classifiers – AQ20, C4.5, C4.5rules CN2 and RIPPER – were employed to classify simulated networks as random or scalefree. In the best case, an average accuracy of over 96% could be achieved by C4.5rules in ten crossvalidation runs.