k-Shell decomposition reveals structural properties of the gene coexpression network for neurodevelopment

Abstract

Neurodevelopment is a dynamic and complex process, which involves interactions of thousands of genes. Understanding the mechanisms of brain development is important for uncovering the genetic architectures of neurodevelopmental disorders such as autism spectrum disorder and intellectual disability. The BrainSpan dataset is an important resource for studying the transcriptional mechanisms governing neurodevelopment. It contains RNA-seq and microarray data for 13 developmental periods in 8-16 brain regions. Various important studies used this dataset, in particular to generate gene coexpression networks. The topology of the BrainSpan gene coexpression network yielded various important gene clusters, which are found to play key roles in diseases. In this work, we analyze the topology of the BrainSpan gene coexpression network using the k-shell decomposition method. k-Shell decomposition is an unsupervised method to decompose a network into layers (shells) using the connectivity information and to detect a nucleus that is central to overall connectivity. Our results show that there are 267 layers in the BrainSpan gene coexpression network. The nucleus contains 2584 genes, which are related to chromatin modification function. We compared and contrasted the structure with the autonomous system level Internet. We found that despite similarities in percolation transition and crust size distribution, there are also differences: the BrainSpan coexpression network has a significantly large nucleus and only a very small number of genes need to access the nucleus first, to be able to connect to other genes in the crust above the nucleus. © TÜBİTAK.