Co-expression networks reveal the tissue-specific regulation of transcription and splicing

Abstract

Gene co-expression networks capture biologically important patterns in gene expression data, enabling functional analyses of genes, discovery of biomarkers, and interpretation of regulatory genetic variants. Most network analyses to date have been limited to assessing correlation between total gene expression levels in a single or small sets of tissues. Here, we have reconstructed networks that capture a much more complete set of regulatory relationships, specifically including regulation of relative isoform abundance and splicing, and tissue-specific connections unique to each of a diverse set of tissues. Using the Genotype-Tissue Expression (GTEx) project v6 RNA-sequencing data across 44 tissues in 449 individuals, we evaluated shared and tissue-specific network relationships. First, we developed a framework called Transcriptome Wide Networks (TWNs) for combining total expression and relative isoform levels into a single sparse network, capturing the complex interplay between the regulation of splicing and transcription. We built TWNs for sixteen tissues, and found that hubs with isoform node neighbors in these networks were strongly enriched for splicing and RNA binding genes, demonstrating their utility in unraveling regulation of splicing in the human transcriptome, and providing a set of candidate shared and tissue-specific regulatory hub genes. Next, we used a Bayesian biclustering model that identifies network edges between genes with co-expression in a single tissue to reconstruct tissue-specific networks (TSNs) for 27 distinct GTEx tissues and for four 1 . CC-BY-NC-ND4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/078741 doi: bioRxiv preprint first posted online Oct. 2, 2016;subsets of related tissues. Using both TWNs and TSNs, we characterized gene co-expression patterns shared across tissues. Finally, we found genetic variants associated with multiple neighboring nodes in our networks, supporting the estimated network structures and identifying 33 genetic variants with distant regulatory impact on transcription and splicing. Our networks provide an improved understanding of the complex relationships between genes in the human transcriptome, including tissue-specificity of gene co-expression, regulation of splicing, and the coordinated impact of genetic variation on transcription.

Regulation of alternative mRNA splicing: old players and new perspectives

“…A). This distinction is however empirical, and whole‐transcriptome sequencing data from a growing number of diverse tissue types suggests that the range of alternatively spliced human transcripts may be greater than previously appreciated. For most categories of splicing events, multiple splice site will be simultaneous present on the nascent pre‐mRNA.…”

“…The mRNA isoform composition varies profoundly across human tissues . It is therefore not surprising that some splicing regulators display tissue‐dependent expression patterns , resulting in corresponding cell type‐specific regulation of their target pre‐mRNA substrates.…”

Environmental perturbations lead to extensive directional shifts in RNA processing

“…Further, they showed that splicing QTLs are also enriched for genetic variants associated with several complex traits in Genome-Wide Association Studies (GWAS), demonstrating the potential importance of splicing misregulation in complex traits [22]. Previous work from our lab and others have shown that gene-by-environment interactions can impact both gene expression and complex traits [23,24,25,26,27,28]. While splicing QTLs have been identified both in humans and mice [22,29,30,31], less is known about how gene-by-environment interactions may affect RNA processing.…”

SoyCSN: Soybean context‐specific network analysis and prediction based on tissue‐specific transcriptome data

“…These hubs usually have important biological functions and generally play vital roles in the network. Based on the size of the network and analogous work in human studies (Saha et al, ), we define three types of hubs by their degrees, small hubs with degrees between 30 and 50, medium hubs with degrees between 50 and 80, and large hubs of degrees more than 80. A summary of nodes, edges, average degree per node, and the number of the hubs is shown in Table .…”

“…Recently, Gao et al developed BicMix, a Bayesian biclustering method, to construct differential gene co‐expression networks (Gao, McDowell, Zhao, Brown, & Engelhardt, ). Saha et al () used this method to construct co‐expression networks on GTEx data across human tissues to study tissue‐specific regulation of transcription and splicing. Xiao, Moreno‐Moral, Rotival, Bottolo, and Petretto () developed a higher‐order generalized singular value decomposition method on multi‐tissue analysis of co‐expression networks.…”