CNET: an algorithm for the inference of gene regulatory interactions from gene expression time series.

The process of inferring regulatory interactions among genes from DNA microarray experiments is known as Reverse Engineering. We present a novel reverse engineering algorithm, CNET, which exploits the principles of Shannon Entropy and Mutual Information through a heuristic scoring function. This function is designed to discover causal relations even if gene temporal profiles exhibit non ideal behaviours, such as noise, quantization errors and variable regulatory delays. Experimental results, both on simulated and on real datasets, show that CNET achieves performance comparable to the state of the art methods in reverse engineering and doubles its performance when inferring on regulatory effects directly dependent on a perturbed target.