Reconstructing Directed Signed Gene Regulatory Network From Microarray Data

Great efforts have been made to develop both algorithms that reconstruct gene regulatory networks and systems that simulate gene networks and expression data, for the purpose of benchmarking network reconstruction algorithms. An interesting observation is that although many simulation systems chose to use Hill kinetics to generate data, none of the reconstruction algorithms were developed based on the Hill kinetics. One possible explanation is that, in Hill kinetics, activation and inhibition interactions take different mathematical forms, which brings additional combinatorial complexity into the reconstruction problem. We propose a new model that qualitatively behaves similar to the Hill kinetics, but has the same mathematical form for both activation and inhibition. We developed an algorithm to reconstruct gene networks based on this new model. Simulation results suggested a novel biological hypothesis that in gene knockout experiments, repressing protein synthesis to a certain extent may lead to better expression data and higher network reconstruction accuracy.