The functions of specialized metabolic genes remain difficult to elucidate, as they often belong to large gene families that undergo frequent gene duplication. While the most common method of gene function prediction -- the transfer of known gene functions through related species -- is easily implementable, it may yield inaccurate conclusions for metabolic genes. As diverse agronomic traits are controlled by metabolic pathways, the scarcity of functionally annotated metabolic genes has wide-reaching implications in agriculture and the plant sciences.   In my talk, I will describe the results of an integrated transcriptomics and metabolomics experiment in the model C3 species Brachypodium distachyon, designed to elucidate metabolic functions of genes. B. distachyon plants were grown under different conditions (heat, Copper [Cu] deficiency, combined heat-Cu deficiency stress, low Phosphate [P], and Arbuscular Mycorrhizal Symbiosis [AMS]). Transcripts and metabolites were simultaneously extracted from leaves and roots, and RNA-seq and untargeted metabolomics was performed. An Information theoretic analysis of the metabolomics data revealed that while leaves had the most complex metabolite profiles, root metabolomes were more inducible and produced more condition-specific metabolites. Differential analysis of metabolite levels coupled with deep learning techniques helped identify broadly stress-responsive metabolite classes. Many condition-specific metabolites – such as blumenols (C13 apocarotenoids) – that can potentially serve as stress biomarkers were also found. We show that while emerging algorithms for metabolite annotation can dramatically increase the interpretability of untargeted metabolomics data, the accuracy of these algorithms should be verified beforehand on in-house, gold standard data. After determining gene expression, we quantified the correlation of both experimentally validated and predicted pathways. Next, we identified genes with conserved expression under AMS and with structurally similar, highly correlated metabolites as candidates of AMS-regulated biosynthetic pathways. This workflow successfully identified both known genes, and new candidate genes regulating plant metabolism under AMS in B. distachyon. Associating stress-induced genes with metabolic pathways provides i) an alternative and effective method of gene function prediction in understudied species, and ii) gene targets for the breeding of more stress-tolerant plants.