Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Sustainable Perennial Crops Laboratory » Research » Publications at this Location » Publication #423500
Research Project: Development of Pathogen- and Plant-Based Genetic Tools and Disease Mitigation Methods for Tropical Perennial Crops

Location: Sustainable Perennial Crops Laboratory

Title: Machine learning reveals complex genetics of fungal resistance in sorghum grain mold

Author
item Ahn, Ezekiel
item Prom, Louis
item Park, Sunchung
item LEE, DONGHO - Orise Fellow
item BHATT, JISHNU - Orise Fellow
item ELLUR, VISHNUTEJ - Washington State University
item LIM, SEUNGHYUN - Orise Fellow
item Jang, Jae Hee
item Lakshman, Dilip
item MAGILL, CLINT - West Texas A & M University

Submitted to: Heredity
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/6/2025
Publication Date: 7/19/2025
Citation: Ahn, E.J., Prom, L.K., Park, S., Lee, D., Bhatt, J., Ellur, V., Lim, S., Jang, J., Lakshman, D.K., Magill, C. 2025. Machine learning reveals complex genetics of fungal resistance in sorghum grain mold. Heredity. https://doi.org/10.1038/s41437-025-00783-9.
DOI: https://doi.org/10.1038/s41437-025-00783-9

Interpretive Summary: Sorghum, a staple crop for millions worldwide, is under constant threat from grain mold disease, significantly reducing yields and food security. Developing sorghum varieties resistant to this disease is crucial, but the complex genetics of resistance makes breeding challenging. In this study, we used powerful machine learning techniques to analyze the genetic makeup of diverse sorghum lines and pinpoint genes associated with grain mold resistance. By comparing how plants responded to disease inoculation versus control conditions, we uncovered a network of genes contributing to resistance. Our findings highlight the intricate genetic nature of grain mold resistance in sorghum, involving many genes working together. We identified promising candidate genes, including those involved in DNA repair, pathogen recognition, and plant defense signaling. This research provides breeders with valuable information and genetic markers to develop sorghum varieties with improved and potentially more durable resistance to grain mold. Ultimately, this work contributes to securing sorghum harvests and ensuring food supplies for communities that rely on this vital crop.

Technical Abstract: This study employed machine learning-driven genome-wide association studies (GWAS) to dissect the complex genetic architecture of grain mold resistance in sorghum (Sorghum bicolor). We utilized phenotypic data from a diverse panel of 306 sorghum accessions, evaluated under inoculation with Alternaria alternata, a mixture of grain mold pathogens, and control conditions. A 'difference phenotype' approach (treatment minus control) was integrated with Boosted Tree and Bootstrap Forest machine learning models to identify significant SNP-trait associations across raw treatment phenotypes, difference phenotypes, a combined phenotype, and PC1. Our analyses confirmed the polygenic nature of grain mold resistance and revealed numerous candidate SNPs across the genome. Consistently identified SNPs were located near genes including Sobic.005G141700 (encoding a protein with DNA repair, pathogen recognition, and defense domains), Sobic.003G329100 (Protein MIZU-KUSSEI 1), and Sobic.002G270800 (PROTEIN ROOT PRIMORDIUM DEFECTIVE 1 - RPD1), each associated with different aspects of resistance. This research demonstrates the efficacy of machine learning-enhanced GWAS, particularly with difference phenotype analysis, for dissecting complex disease resistance. The identified candidate genes and robust SNP markers provide valuable resources for marker-assisted selection and genomic selection to accelerate breeding efforts for grain mold-resistant sorghum varieties.