Integrating GWAS and gene coexpression networks to identify biotic stress resistance genes in soybean (Glycine max)

Soybean (Glycine max (L.) Merr.) is the most important legume crop worldwide, and it is important for food, feed and industrial applications. Diseases and pests are responsible for significant yield losses for soybean farmers around the world. Several association studies have identified biotic stress resistance markers in the soybean genome, but identifying the causal genes remains a challenge. Here, we integrated public GWAS and transcriptome data sets to reconstruct gene coexpression networks and identify high confidence candidate genes involved in biotic stress resistance. By manually curating the scientific literature, we identified 685 markers related to disease and pest resistance, of which 293 are associated to fungi, 154 to oomycetes, 110 to insects, 106 to nematodes and 26 to viruses. By using the R packages GenomicRanges and IRanges, we extracted all genes within a 1 Mb-sliding window relative to each marker, which resulted in 17375, 9299, 8994, 5644 and 931 candidate resistance genes for fungi, nematodes, oomycetes, insects and viruses, respectively. Transcriptome data were downloaded from the database Soybean Expression Atlas. Experimentally validated resistance genes and genes encoding pathogenesis-related (PR) proteins were downloaded from UniProt and used as guide genes. For each data set, we removed genes with median values below 1 and selected the top 80% most variable genes. The R package BioNERO was used for data pre-processing, quantile normalization and gene coexpression network reconstruction. High confidence candidate genes were defined as the genes that were coexpressed with guide genes and were significantly induced or repressed under stress conditions. The systematic filtering dramatically reduced the list of candidate genes. Functional enrichment analyses revealed that the high confidence candidates are involved in important metabolic pathways and biological processes for disease and pest resistance. These results will help prioritize candidate genes for downstream experiments and indicate promising genes for developing mutant and/or transgenic lines.