Search for: All records

Abiotic stresses such as drought, heat, cold, salinity and flooding significantly impact plant growth, development and productivity. As the planet has warmed, these abiotic stresses have increased in frequency and intensity, affecting the global food supply and making it imperative to develop stress-resilient crops. In the past 20 years, the development of omics technologies has contributed to the growth of datasets for plants grown under a wide range of abiotic environments. Integration of these rapidly growing data using machine-learning (ML) approaches can complement existing breeding efforts by providing insights into the mechanisms underlying plant responses to stressful conditions, which can be used to guide the design of resilient crops. In this review, we introduce ML approaches and provide examples of how researchers use these approaches to predict molecular activities, gene functions and genotype responses under stressful conditions. Finally, we consider the potential and challenges of using such approaches to enable the design of crops that are better suited to a changing environment. This article is part of the theme issue ‘Crops under stress: can we mitigate the impacts of climate change on agriculture and launch the ‘Resilience Revolution’?’. 

The formation of complex traits is the consequence of genotype and activities at multiple molecular levels. However, connecting genotypes and these activities to complex traits remains challenging. Here, we investigate whether integrating genomic, transcriptomic, and methylomic data can improve pre- diction for six Arabidopsis traits. We find that transcriptome- and methylome- based models have performances comparable to those of genome-based models. However, models built for flowering time using different omics data identify different benchmark genes. Nine additional genes identified as important for flowering time from our models are experimentally validated as regulating flowering. Gene contributions to flowering time prediction are accession-dependent and distinct genes contribute to trait prediction in dif- ferent genotypes. Models integrating multi-omics data perform best and reveal known and additional gene interactions, extending knowledge about existing regulatory networks underlying flowering time determination. These results demonstrate the feasibility of revealing molecular mechanisms underlying complex traits through multi-omics data integration. 

Scientific advances due to conceptual or technological innovations can be revealed by examining how research topics have evolved. But such topical evolution is difficult to uncover and quantify because of the large body of literature and the need for expert knowledge in a wide range of areas in a field. Using plant biology as an example, we used machine learning and language models to classify plant science citations into topics representing interconnected, evolving subfields. The changes in prevalence of topical records over the last 50 years reflect shifts in major research trends and recent radiation of new topics, as well as turnover of model species and vastly different plant science research trajectories among countries. Our approaches readily summarize the topical diversity and evolution of a scientific field with hundreds of thousands of relevant papers, and they can be applied broadly to other fields. 

Abstract Natural language processing (NLP) techniques can enhance our ability to interpret plant science literature. Many state-of-the-art algorithms for NLP tasks require high-quality labelled data in the target domain, in which entities like genes and proteins, as well as the relationships between entities, are labelled according to a set of annotation guidelines. While there exist such datasets for other domains, these resources need development in the plant sciences. Here, we present the Plant ScIenCe KnowLedgE Graph (PICKLE) corpus, a collection of 250 plant science abstracts annotated with entities and relations, along with its annotation guidelines. The annotation guidelines were refined by iterative rounds of overlapping annotations, in which inter-annotator agreement was leveraged to improve the guidelines. To demonstrate PICKLE’s utility, we evaluated the performance of pretrained models from other domains and trained a new, PICKLE-based model for entity and relation extraction (RE). The PICKLE-trained models exhibit the second-highest in-domain entity performance of all models evaluated, as well as a RE performance that is on par with other models. Additionally, we found that computer science-domain models outperformed models trained on a biomedical corpus (GENIA) in entity extraction, which was unexpected given the intuition that biomedical literature is more similar to PICKLE than computer science. Upon further exploration, we established that the inclusion of new types on which the models were not trained substantially impacts performance. The PICKLE corpus is, therefore, an important contribution to training resources for entity and RE in the plant sciences. 

Abstract Genetic redundancy refers to a situation where an individual with a loss-of-function mutation in one gene (single mutant) does not show an apparent phenotype until one or more paralogs are also knocked out (double/higher-order mutant). Previous studies have identified some characteristics common among redundant gene pairs, but a predictive model of genetic redundancy incorporating a wide variety of features derived from accumulating omics and mutant phenotype data is yet to be established. In addition, the relative importance of these features for genetic redundancy remains largely unclear. Here, we establish machine learning models for predicting whether a gene pair is likely redundant or not in the model plant Arabidopsis thaliana based on six feature categories: functional annotations, evolutionary conservation including duplication patterns and mechanisms, epigenetic marks, protein properties including post-translational modifications, gene expression, and gene network properties. The definition of redundancy, data transformations, feature subsets, and machine learning algorithms used significantly affected model performance based on hold-out, testing phenotype data. Among the most important features in predicting gene pairs as redundant were having a paralog(s) from recent duplication events, annotation as a transcription factor, downregulation during stress conditions, and having similar expression patterns under stress conditions. We also explored the potential reasons underlying mispredictions and limitations of our studies. This genetic redundancy model sheds light on characteristics that may contribute to long-term maintenance of paralogs, and will ultimately allow for more targeted generation of functionally informative double mutants, advancing functional genomic studies. 

Abstract Plant specialized metabolites mediate interactions between plants and the environment and have significant agronomical/pharmaceutical value. Most genes involved in specialized metabolism (SM) are unknown because of the large number of metabolites and the challenge in differentiating SM genes from general metabolism (GM) genes. Plant models like Arabidopsis thaliana have extensive, experimentally derived annotations, whereas many non-model species do not. Here we employed a machine learning strategy, transfer learning, where knowledge from A. thaliana is transferred to predict gene functions in cultivated tomato with fewer experimentally annotated genes. The first tomato SM/GM prediction model using only tomato data performs well (F-measure = 0.74, compared with 0.5 for random and 1.0 for perfect predictions), but from manually curating 88 SM/GM genes, we found many mis-predicted entries were likely mis-annotated. When the SM/GM prediction models built with A. thaliana data were used to filter out genes where the A. thaliana-based model predictions disagreed with tomato annotations, the new tomato model trained with filtered data improved significantly (F-measure = 0.92). Our study demonstrates that SM/GM genes can be better predicted by leveraging cross-species information. Additionally, our findings provide an example for transfer learning in genomics where knowledge can be transferred from an information-rich species to an information-poor one. 

Plant specialized metabolism (SM) enzymes produce lineage-specific metabolites with important ecological, evolutionary, and biotechnological implications. UsingArabidopsis thalianaas a model, we identified distinguishing characteristics of SM and GM (general metabolism, traditionally referred to as primary metabolism) genes through a detailed study of features including duplication pattern, sequence conservation, transcription, protein domain content, and gene network properties. Analysis of multiple sets of benchmark genes revealed that SM genes tend to be tandemly duplicated, coexpressed with their paralogs, narrowly expressed at lower levels, less conserved, and less well connected in gene networks relative to GM genes. Although the values of each of these features significantly differed between SM and GM genes, any single feature was ineffective at predicting SM from GM genes. Using machine learning methods to integrate all features, a prediction model was established with a true positive rate of 87% and a true negative rate of 71%. In addition, 86% of known SM genes not used to create the machine learning model were predicted. We also demonstrated that the model could be further improved when we distinguished between SM, GM, and junction genes responsible for reactions shared by SM and GM pathways, indicating that topological considerations may further improve the SM prediction model. Application of the prediction model led to the identification of 1,220A. thalianagenes with previously unknown functions, each assigned a confidence measure called an SM score, providing a global estimate of SM gene content in a plant genome. 

Summary Plant metabolites from diverse pathways are important for plant survival, human nutrition and medicine. The pathway memberships of most plant enzyme genes are unknown. While co‐expression is useful for assigning genes to pathways, expression correlation may exist only under specific spatiotemporal and conditional contexts.Utilising > 600 tomato (Solanum lycopersicum) expression data combinations, three strategies for predicting memberships in 85 pathways were explored.Optimal predictions for different pathways require distinct data combinations indicative of pathway functions. Naive prediction (i.e. identifying pathways with the most similarly expressed genes) is error prone. In 52 pathways, unsupervised learning performed better than supervised approaches, possibly due to limited training data availability. Using gene‐to‐pathway expression similarities led to prediction models that outperformed those based simply on expression levels. Using 36 experimental validated genes, the pathway‐best model prediction accuracy is 58.3%, significantly better compared with that for predicting annotated genes without experimental evidence (37.0%) or random guess (1.2%), demonstrating the importance of data quality.Our study highlights the need to extensively explore expression‐based features and prediction strategies to maximise the accuracy of metabolic pathway membership assignment. The prediction framework outlined here can be applied to other species and serves as a baseline model for future comparisons.