Integrative metabolome-genome analysis reveals the genetic architecture of metabolic diversity in sorghum grain

Cereal grains are fundamental to global food security and bioenergy production, yet the genetic and molecular bases of grain metabolic diversity remain unresolved. Sorghum, a climate-resilient C4 grass with exceptional tolerance to heat and drought stress, provides an effective system for metabolic genomics. Here, we integrated large-scale untargeted metabolomic profiling, population genomics, and machine learning to systematically dissect grain metabolic diversity and its genetic determinants in sorghum. Untargeted metabolomic analysis of mature grains from the Sorghum Association Panel detected 4,877 metabolites exhibiting extensive quantitative variation across accessions, with pronounced diversity in amino acid and polyphenol biosynthetic pathways, indicating substantial standing variation relevant to nutritional improvement. Metabolite-based genome-wide association studies (mGWAS) identified [~]4.15 million significant SNP-metabolite associations, significantly enriched in noncoding regulatory regions, consistent with a highly polygenic architecture of metabolic regulation. We further delineated 38 metabolite-gene clusters that reveal coordinated genetic control of core metabolic pathways, and used machine learning to pinpoint key metabolites underlying grain color variation and to prioritize associated candidate genes, demonstrating predictive models that integrate genotype, metabolome, and phenotype. This work establishes the first population-scale atlas of sorghum grain metabolomic and genetic diversity and introduces the Sorghum Grain Metabolite Diversity Atlas (SorGMDA), an open resource integrating metabolomic and genomic variation. These resources enable comparative metabolic genomics across cereals and support systems-level breeding strategies for improving grain nutritional quality and climate resilience.