Construction of Malaria Gene Expression Network Using Partial Correlations

In this paper we model the gene expression network of Plasmodium falciparum using the time-course microarray dataset. A gene expression network is constructed based on a novel method that combines two types of correlations between each pair of genes: standard Pearson and partial correlations. A link is established between two genes if both correlation coefficients are higher than their corresponding thresholds. The values for thresholds are sought so that the topology of the resulting network satisfies several criteria. The sought network has to be sparse, small-world (with any two genes being connected by a path of a few links only), scale-free-like (wherein a small number of genes have a large number of links and many genes have only a few connections). Similar to gene networks of other organisms the highly connected genes (hubs) in the constructed network tend to have essential cell functions. To verify the proposed method and to compare the results, a scale-free-like, small-world gene expression network was also constructed using another dataset, confirming the lethality and centrality property of malaria hubs.