Improving gene regulatory network inference and assessment: The importance of using network structure

Abstract

Gene regulatory networks are graph models representing cellular transcription events. Networks are far from complete due to time and resource consumption for experimental validation and curation of the interactions. Previous assessments have shown the modest performance of the available network inference methods based on gene expression data. Here, we study several caveats on the inference of regulatory networks and methods assessment through the quality of the input data and gold standard, and the assessment approach with a focus on the global structure of the network. We used synthetic and biological data for the predictions and experimentally-validated biological networks as the gold standard (ground truth). Standard performance metrics and graph structural properties suggest that methods inferring co-expression networks should no longer be assessed equally with those inferring regulatory interactions. While methods inferring regulatory interactions perform better in global regulatory network inference than co-expression-based methods, the latter is better suited to infer function-specific regulons and co-regulation networks. When merging expression data, the size increase should outweigh the noise inclusion and graph structure should be considered when integrating the inferences. We conclude with guidelines to take advantage of inference methods and their assessment based on the applications and available expression datasets.

Network analyses reveal D clade ethylene response factors as major regulators of jasmonic acid-mediated resistance to early blight disease complex in tomato

“…To identify the genes with the strongest influence on the topology of the defense-related co-expression network, we calculated network eigenvetor centrality for each gene in a given co-expression module. These genes are central to the directed biological specificity of their local co-expression modules [79]. We found 20, 53, and 94 genes with high eigenvector centrality for the yellow, turquoise, and blue co-expression modules respectively ( supp.…”

“…To identify global regulators of the GRN, we filtered the global GRN for co-expression module membership, and calculated network eigenvector centrality for each gene in a given GRN subnetwork. While these genes may not have a direct role in the biological activity of the co-expression module, but they regulate the regulators due to their basal role in the global regulatory hierarchy [79]. We calculated 40 hub genes within the blue module GRN ( supp.…”

Unraveling patient heterogeneity in complex diseases through individualized co-expression networks: a perspective

“…Integrating multi-omics data from different modalities (e.g., genomics, transcriptomics, proteomics) could provide a more comprehensive view of individual-specific networks. Collaboration among researchers and data sharing can help increase sample sizes and improve the statistical power of coexpression network inference ( Escorcia-Rodríguez et al, 2023 ). The development of novel statistical methods specifically designed for analyzing individualized coexpression networks can improve the accuracy and reliability of the inferred networks ( Yu et al, 2018 ).…”

“…However, there are still challenges to overcome. When merging expression data, the size increase should outweigh the noise inclusion, and graph structure should be considered when integrating the inferences ( Escorcia-Rodríguez et al, 2023 ). The potential bias introduced by relying on external datasets should also be considered, as they may only partially represent the specific biological context of the individual sample.…”

Topological benchmarking of algorithms to infer gene regulatory networks from single-cell RNA-seq data

“…The structural properties of networks can be quantified by topological measurements ( Koutrouli et al 2020 ), including, for instance, the network efficiency and the assortativity. So far, the performance of GRN inference algorithms on the estimation of topological properties has only been assessed with bulk RNA-seq data ( Kiani et al 2016 , Escorcia-Rodríguez et al 2023 ), and by employing a limited number of synthetic networks ( Kiani et al 2016 ), which makes it hard to reach robust conclusions for single-cell data.…”