Information theory and machine learning illuminate large-scale metabolomic responses of Brachypodium distachyon to environmental change

Plant responses to environmental change are mediated via changes in cellular metabolomes. However, <5% of signals obtained from liquid chromatography tandem mass spectrometry (LC-MS/MS) can be identified, limiting our understanding of how metabolomes change under biotic/abiotic stress. To address this challenge, we performed untargeted LC-MS/MS of leaves, roots, and other organs of Brachypodium distachyon (Poaceae) under 17 organ-condition combinations, including copper deficiency, heat stress, low phosphate, and arbuscular mycorrhizal symbiosis. We found that both leaf and root metabolomes were significantly affected by the growth medium. Leaf metabolomes were more diverse than root metabolomes, but the latter were more specialized and more responsive to environmental change. We found that 1 week of copper deficiency shielded the root, but not the leaf metabolome, from perturbation due to heat stress. Machine learning (ML)-based analysis annotated approximately 81% of the fragmented peaks versus approximately 6% using spectral matches alone. We performed one of the most extensive validations of ML-based peak annotations in plants using thousands of authentic standards, and analyzed approximately 37% of the annotated peaks based on these assessments. Analyzing responsiveness of each predicted metabolite class to environmental change revealed significant perturbations of glycerophospholipids, sphingolipids, and flavonoids. Co-accumulation analysis further identified condition-specific biomarkers. To make these results accessible, we developed a visualization platform on the Bio-Analytic Resource for Plant Biology website (https://bar.utoronto.ca/efp_brachypodium_metabolites/cgi-bin/efpWeb.cgi), where perturbed metabolite classes can be readily visualized. Overall, our study illustrates how emerging chemoinformatic methods can be applied to reveal novel insights into the dynamic plant metabolome and stress adaptation.

Illuminating the lineage-specific diversification of resin glycoside acylsugars in the morning glory (Convolvulaceae) family using computational metabolomics

Acylsugars are a class of plant defense compounds produced across many distantly related families. Members of the horticulturally important morning glory (Convolvulaceae) family produce a diverse sub-class of acylsugars called resin glycosides (RGs), which comprise oligosaccharide cores, hydroxyacyl chain(s), and decorating aliphatic and aromatic acyl chains. While many RG structures are characterized, the extent of structural diversity of this class in different genera and species is not known. In this study, we asked whether there has been lineage-specific diversification of RG structures in different Convolvulaceae species that may suggest diversification of the underlying biosynthetic pathways. Liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS) was performed from root and leaf extracts of 26 species sampled in a phylogeny-guided manner. LC-MS/MS revealed thousands of peaks with signature RG fragmentation patterns with one species producing over 300 signals, mirroring the diversity in Solanaceae-type acylsugars. A novel RG from Dichondra argentea was characterized using Nuclear Magnetic Resonance spectroscopy, supporting previous observations of RGs with open hydroxyacyl chains instead of closed macrolactone ring structures. Substantial lineage-specific differentiation in utilization of sugars, hydroxyacyl chains, and decorating acyl chains was discovered, especially among Ipomoea and Convolvulus - the two largest genera in Convolvulaceae. Adopting a computational, knowledge-based strategy, we further developed a high-recall workflow that successfully explained ~72% of the MS/MS fragments, predicted the structural components of 11/13 previously characterized RGs, and partially annotated ~45% of the RGs. Overall, this study improves our understanding of phytochemical diversity and lays a foundation for characterizing the evolutionary mechanisms underlying RG diversification.

Machine learning: A powerful tool for gene function prediction in plants

Recent advances in sequencing and informatic technologies have led to a deluge of publicly available genomic data. While it is now relatively easy to sequence, assemble, and identify genic regions in diploid plant genomes, functional annotation of these genes is still a challenge. Over the past decade, there has been a steady increase in studies utilizing machine learning algorithms for various aspects of functional prediction, because these algorithms are able to integrate large amounts of heterogeneous data and detect patterns inconspicuous through rule-based approaches. The goal of this review is to introduce experimental plant biologists to machine learning, by describing how it is currently being used in gene function prediction to gain novel biological insights. In this review, we discuss specific applications of machine learning in identifying structural features in sequenced genomes, predicting interactions between different cellular components, and predicting gene function and organismal phenotypes. Finally, we also propose strategies for stimulating functional discovery using machine learning-based approaches in plants.