Inference of gene regulatory networks from large scale gene expression data

With the advent of the age of genomics, an increasing number of genes have been identified and their functions documented. However, not as much is known of specific regulatory relations among genes (e.g. gene A up-regulates gene B). At the same time, there is an increasing number of large-scale gene expression datasets, in which the mRNA transcript levels of tens of thousands of genes are measured at a number of time points, or under a number of different conditions. A number of studies have proposed to find gene regulatory networks from such datasets. Our method is a modification of the continuous-time neural network method of Wahde & Hertz [25, 26]. The genetic algorithm used to update weights was replaced with Levenberg-Marquardt optimization. We tested our method on artificial data as well as Spellman's yeast cell cycle data [22]. Results indicated that this method was able to detect salient regulatory relations between genes