Deep Learning Genome-wide Linkage Association Study for Wheat Fusarium Head Blight Resistance Genes Discovery

Abstract

Background: Fusarium head blight (FHB) is one of the most devastating diseases of wheat worldwide and artificial intelligence can assist with understanding resistance to the disease. Considering different sample populations, marker types, reference maps, and statistical methods, we developed a Deep Learning Genome-wide Linkage Association Study (dpGLAS) of FHB resistance in wheat. Results: The dpGLAS was first applied to two bi-parental population datasets in which the cultivar AC Barrie was a common parent for FHB resistance. Eight candidate gene markers were discovered in the one AC Barrie population and 10 in the other associated with FHB resistance. Eight of these markers were also supported by the conventional QTL mapping. Most of these candidate marker genes were found associated with the Reactive Oxygen Species (ROS) and Abscisic acid (ABA) axes. These ROS and ABA pathways were further supported by RNA-seq transcriptome data of FHB resistant cv. AAC Tenacious, a parent of the third bi-parental population. In this dataset, the ROS-centered Panther protein families were significantly enriched in those genes that had most different response to FHB when compared the resistance Tenacious and the susceptible Roblin. Conclusions: This study developed the framework of dpGLAS and identified candidate genes for FHB resistance in the Canadian spring wheat cultivars AC Barrie and AAC Tenacious.