Bridging metabolomics with plant phenotypic responses is challenging. Multivariate analyses account for the existing dependencies among metabolites, and regression models in particular capture such dependencies in search for association with a given trait. However, special care should be undertaken with metabolomics data. Here we propose a modeling workflow that considers all caveats imposed by such large data sets.
Keywords: Metabolomics; Modeling; Plants; Prediction; R programing language; R software packages; Regression; Trait.