Inferring Gene Regulatory Network Structure

Inferring the network structure of gene regulatory networks is one of the most important problems in contemporary bioinformatics. We analyze different methodologies for inferring small to very large sized gene networks. We use the datasets of DREAM 3 in-silico network challenge that is provided online [1]. The challenge involves inferring primarily the network structure from steady state gene expressions, knockout and knockdown and time-series data for gene networks of size 10, 50 and 100. The basis for all such inference methodologies in gene networks is the assumption that the network is sparse [2],[3]. We look at three formulations that involve time series data and knockout data. We propose a heuristic for inference from knock out data only that provides reasonably good results on the datasets. We also compare the performance of our heuristic with that of other top teams and the results indicate that our proposed methodology fares in the top three rankings consistently.