Michno, J.-M. | bioRxiv Science

bioRxiv2018

Identification of candidate genes underlying nodulation-specific phenotypes in Medicago truncatula through integration of genome-wide association studies and co-expression networks

Genome-wide association studies (GWAS) have proven to be a valuable approach for identifying genetic intervals associated with phenotypic variation in Medicago truncatula. These intervals can vary in size, depending on the historical local recombination near each significant interval. Typically, significant intervals span numerous gene models, limiting the ability to resolve high-confidence candidate genes underlying the trait of interest. Additional genomic data, including gene co-expression networks, can be combined with the genetic mapping information to successfully identify candidate genes. Co-expression network analysis provides information about the functional relationships of each gene through its similarity of expression patterns to other well-defined clusters of genes. In this study, we integrated data from GWAS and co-expression networks to pinpoint candidate genes that may be associated with nodule-related phenotypes in Medicago truncatula. We further investigated a subset of these genes and confirmed that several had existing evidence linking them nodulation, including MEDTR2G101090 (PEN3-like), a previously validated gene associated with nodule number.

systems biology↗

bioRxiv2017

Integrating co-expression networks with GWAS detects genes driving elemental accumulation in maize seeds

BackgroundGenome wide association studies (GWAS) have identified thousands of loci linked to hundreds of traits in many different species. However, because linkage equilibrium implicates a broad region surrounding each identified locus, the causal genes often remain unknown. This problem is especially pronounced in non-human, non-model species where functional annotations are sparse and there is frequently little information available for prioritizing candidate genes.\n\nResultsTo address this issue, we developed a computational approach called Camoco (Co-Analysis of Molecular Components) that systematically integrates loci identified by GWAS with gene co-expression networks to prioritize putative causal genes. We applied Camoco to prioritize candidate genes from a large-scale GWAS examining the accumulation of 17 different elements in maize seeds. Camoco identified statistically significant subnetworks for the majority of traits examined, producing a prioritized list of high-confidence causal genes for several agronomically important maize traits. Two candidate genes identified by our approach were validated through analysis of mutant phenotypes. Strikingly, we observed a strong dependence in the performance of our approach on the type of co-expression network used: expression variation across genetically diverse individuals in a relevant tissue context (in our case, maize roots) outperformed other alternatives.\n\nConclusionsOur study demonstrates that co-expression networks can provide a powerful basis for prioritizing candidate causal genes from GWAS loci, but suggests that the success of such strategies can highly depend on the gene expression data context. Both the Camoco software and the lessons on integrating GWAS data with co-expression networks generalize to species beyond maize.

systems biology↗