First large-scale peach gene coexpression network: A new tool for predicting gene function

Transcriptomics studies generate enormous amounts of biological information. Nowadays, representing this complex data as gene coexpression networks (GCNs) is becoming commonplace. Peach is a model for Prunus genetics and genomics, but identifying and validating genes associated to peach breeding traits is a complex task. A GCN capable of capturing stable gene-gene relationships would help researchers overcome the intrinsic limitations of peach genetics and genomics approaches and outline future research opportunities. In this study, we created the first large-scale GCN in peach, applying aggregated and non-aggregated methods to create four GCNs from 604 Illumina RNA-Seq libraries. We evaluated the performance of every GCN in predicting functional annotations using a machine-learning algorithm based on the guilty-by-association principle. The GCN with the best performance was COO300, encompassing 21,956 genes and an average AUROC of 0.746. To validate its performance predicting gene function, we used two well-characterized genes involved in fruit flesh softening in peach: the endopolygalacturonases PpPG21 and PpPG22. Genes coexpressing with PpPG21 and PpPG22 were extracted and named as melting flesh (MF) subnetwork. Finally, we performed an enrichment analysis of MF subnetwork and compared the results with the current knowledge regarding peach fruit softening process. The MF subnetwork mainly included genes involved in cell wall expansion and remodeling, with expression triggered by ripening-related phytohormones such as ethylene, auxin and methyl jasmonates. All these processes are closely related with peach fruit softening and therefore related to the function of PpPG21 and PpPG22. These results validate COO300 as a powerful tool for peach and Prunus research. COO300, renamed as PeachGCN v1.0, and the scripts necessary to perform a function prediction analysis using it, are available at https://github.com/felipecobos/PeachGCN.