Figure 1From: Exploring molecular backgrounds of quality traits in rice by predictive models based on high-coverage metabolomicsMetabolomics characterization of the RDRS. Seeds were collected from field-grown rice and analyzed on 4 metabolomics platforms (a). Multi-platform metabolite profiles were summarized to obtain non-redundant data (b). Quantitative quality trait data were gathered and pre-treated to remove the correlation with genetic population structure (c). MB-OPLS was used to decompose the metabolite profiles to platform-specific systematic bias (d), noise (e) and the trait-correlated variance used for predicting each trait (f). A novel feature selection method was used to identify trait-associated metabolites that were used to generate network visualization (g). Cross-validation and an independent experiment were performed to validate the derived models (h).Back to article page