Rice_MT_HSI_data

Background: Metabolomic analysis is now widely used in crop-related research but can be logistically demanding and costly.Results: We describe a relatively inexpensive approach in which metabolomics is combined with hyperspectral imaging (HSI) via machine learning. This approach was used to target important steps in flavonoid and lipid biosynthesis in rice. We extracted 1848 hyperspectral indices and 887 metabolites from polished grains of 533 O. sativa accessions. Hyperspectral indices were linked to metabolites through correlation analysis and modelling. On this basis, a total of 554 metabolites and 1313 hyperspectral indices were identified for use in a genome-wide association study (GWAS). The GWAS revealed 17,509 significant locus-trait associations with 2882 single nucleotide polymorphism (SNPs). Colocalization analysis linked SNPs to corresponding metabolites and hyperspectral indices, and 6415 pairs of metabolites and hyperspectral indices were associated within a linkage disequilibrium (LD) of 300 kb in the O. sativa genome. Transcriptomic assessments linked the colocalized loci to 1761 candidate genes associated with KEGG pathways. In an important validation step, genes encoding a novel flavonoid (LOC_Os09g18450) and a flavonoid/lipid (LOC_Os07g11020) that were targeted by our method were characterized. LOC_Os07g11020 was colocalized on the basis of hyperspectral indices and metabolite data, whereas LOC_Os09g18450 was colocalized on the basis of the hyperspectral index ddA42 only. The roles of these genes in metabolism were confirmed by gene editing and overexpression in rice lines, which revealed predicted changes in biochemical profiles.Conclusion: Our findings indicate that hyperspectral imaging combined with machine learning methods could serve as a powerful tool for quickly and inexpensively assessing crop metabolites.