Top-Down Analysis of Gene-Regulatory Networks Using Global Coordination Method

Cellular activity is governed by an underlying gene regulatory network. However, due to the large complexity of the regulatory network, inferring them is extremely challenging. A novel computational measure was recently developed to cut across the network of interactions among genes and consider their transcriptional interrelationships in a top-down manner. The new measure, called global coordination level (GCL), is based on the average multivariate dependency between the expression levels of random subsets of genes in single-cell RNA-seq datasets. Yet, there are several fundamental features of this top-down method that need addressing. Here, we systematically analyse the performance and limitations of the GCL using real and simulated gene expression data. We compare the results to other bottom-up methods, such as co-expression matrices and reconstructed networks. We find that the GCL has significant advantage in detecting the gene-to-gene coordination level, especially where the number of available samples is low compared to the number of genes, which is a typical scenario in current available data. The systematic analysis of the GCL provides a useful foundation for future application of this method