Search

New graduate students in biology programs may lack the quantitative skills necessary for their research and professional careers. The acquisition of these skills may be impeded by teaching and mentoring experiences that decrease rather than increase students’ beliefs in their ability to learn and apply quantitative approaches. In this opinion piece, we argue that revising instructional experiences to ensure that both student confidence and quantitative skills are enhanced may improve both educational outcomes and professional success. A few studies suggest that explicitly addressing productive failure in an instructional setting and ensuring effective mentoring may be the most effective routes to simultaneously increasing both quantitative self-efficacy and quantitative skills. However, there is little work that specifically addresses graduate student needs, and more research is required to reach evidence-backed conclusions.

Basic helix–loop–helix (bHLH) proteins are one of the largest families of transcription factor (TF) in eukaryotes, and ~30% of all flowering plants’ bHLH TFs contain the aspartate kinase, chorismate mutase, and TyrA (ACT)-like domain at variable distances C-terminal from the bHLH. However, the evolutionary history and functional consequences of the bHLH/ACT-like domain association remain unknown. Here, we show that this domain association is unique to the plantae kingdom with green algae (chlorophytes) harboring a small number of bHLH genes with variable frequency of ACT-like domain’s presence. bHLH-associated ACT-like domains form a monophyletic group, indicating a common origin. Indeed, phylogenetic analysis results suggest that the association of ACT-like and bHLH domains occurred early in Plantae by recruitment of an ACT-like domain in a common ancestor with widely distributed ACT DOMAIN REPEAT ( ACR ) genes by an ancestral bHLH gene. We determined the functional significance of this association by showing that Chlamydomonas reinhardtii ACT-like domains mediate homodimer formation and negatively affect DNA binding of the associated bHLH domains. We show that, while ACT-like domains have experienced faster selection than the associated bHLH domain, their rates of evolution are strongly and positively correlated, suggesting that the evolution of the ACT-like domains was constrained by the bHLH domains. This study proposes an evolutionary trajectory for the association of ACT-like and bHLH domains with the experimental characterization of the functional consequence in the regulation of plant-specific processes, highlighting the impacts of functional domain coevolution.

Synergid cells in the micropylar end of the female gametophyte are required for critical cell-cell signaling interactions between the pollen tube and the ovule that precede double fertilization and seed formation in flowering plants. LORELEI (LRE) encodes a GPI-anchored protein that is expressed primarily in the synergid cells, and together with FERONIA, a receptor-like kinase, it controls pollen tube reception by the receptive synergid cell. Still, how LRE expression is controlled in synergid cells remains poorly characterized. We identified candidate cis-regulatory elements enriched in LRE and other synergid cell-expressed genes. One of the candidate motifs (‘TAATATCT’) in the LRE promoter was an uncharacterized variant of the Evening Element motif that we named as the Short Evening Element-like (SEEL) motif. Deletion or point mutations in the SEEL motif of the LRE promoter resulted in decreased reporter expression in synergid cells, demonstrating that the SEEL motif is important for expression of LRE in synergid cells. Additionally, we found that LRE expression is decreased in the loss of function mutants of REVEILLE (RVE) transcription factors, which are clock genes known to bind the SEEL and other closely related motifs. We propose that RVE transcription factors regulate LRE expression in synergid cells by binding to the SEEL motif in the LRE promoter. Identification of a cis-regulatory element and transcription factors involved in the expression of LRE will serve as a foundation to characterize the gene regulatory networks in synergid cells and investigate the potential connection between circadian rhythm and fertilization.One sentence summaryA newly identified SEEL motif in the promoter of LORELEI and at least three members of the REVEILLE transcription factor family are important for LORELEI expression in synergid cells of the Arabidopsis female gametophyte.

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.

The superior photosynthetic efficiency of C4leaves over C3leaves is owing to their unique Kranz anatomy, in which the vein is surrounded by one layer of bundle sheath (BS) cells and one layer of mesophyll (M) cells. Kranz anatomy development starts from three contiguous ground meristem (GM) cells, but its regulators and underlying molecular mechanism are largely unknown. To identify the regulators, we obtained the transcriptomes of 11 maize embryonic leaf cell types from five stages of pre-Kranz cells starting from median GM cells and six stages of pre-M cells starting from undifferentiated cells. Principal component and clustering analyses of transcriptomic data revealed rapid pre-Kranz cell differentiation in the first two stages but slow differentiation in the last three stages, suggesting early Kranz cell fate determination. In contrast, pre-M cells exhibit a more prolonged transcriptional differentiation process. Differential gene expression and coexpression analyses identified gene coexpression modules, one of which included 3 auxin transporter and 18 transcription factor (TF) genes, including known regulators of Kranz anatomy and/or vascular development. In situ hybridization of 11 TF genes validated their expression in early Kranz development. We determined the binding motifs of 15 TFs, predicted TF target gene relationships among the 18 TF and 3 auxin transporter genes, and validated 67 predictions by electrophoresis mobility shift assay. From these data, we constructed a gene regulatory network for Kranz development. Our study sheds light on the regulation of early maize leaf development and provides candidate leaf development regulators for future study.

The broad host range of Fusarium virguliforme represents a unique comparative system to identify and define differentially induced responses between an asymptomatic monocot host, maize (Zea mays), and a symptomatic eudicot host, soybean (Glycine max). Using a temporal, comparative transcriptome-based approach, we observed that early gene expression profiles of root tissue from infected maize suggest that pathogen tolerance coincides with the rapid induction of senescence dampening transcriptional regulators, including ANACs (Arabidopsis thaliana NAM/ATAF/CUC protein) and Ethylene-Responsive Factors. In contrast, the expression of senescence-associated processes in soybean was coincident with the appearance of disease symptom development, suggesting pathogen-induced senescence as a key pathway driving pathogen susceptibility in soybean. Based on the analyses described herein, we posit that root senescence is a primary contributing factor underlying colonization and disease progression in symptomatic versus asymptomatic host–fungal interactions. This process also supports the lifestyle and virulence of F. virguliforme during biotrophy to necrotrophy transitions. Further support for this hypothesis lies in comprehensive co-expression and comparative transcriptome analyses, and in total, supports the emerging concept of necrotrophy-activated senescence. We propose that F. virguliforme conditions an environment within symptomatic hosts, which favors susceptibility through transcriptomic reprogramming, and as described herein, the induction of pathways associated with senescence during the necrotrophic stage of fungal development.

Arabidopsis AINTEGUMENTA (ANT), an AP2 transcription factor, is known to control plant growth and floral organogenesis. In this study, our transcriptome analysis and in situ hybridization assays of maize embryonic leaves suggested that maize ANT1 (ZmANT1) regulates vascular development. To better understand ANT1 functions, we determined the binding motif of ZmANT1 and then showed that ZmANT1 binds the promoters of millet SCR1, GNC, and AN3, which are key regulators of Kranz anatomy, chloroplast development, and plant growth, respectively. We generated a mutant with a single-codon deletion and two frameshift mutants of the ANT1 ortholog in the C4 millet Setaria viridis by the CRISPR/Cas9 technique. The two frameshift mutants displayed reduced photosynthesis efficiency and growth rate, smaller leaves, and lower grain yields than wild-type (WT) plants. Moreover, their leaves sporadically exhibited distorted Kranz anatomy and vein spacing. Conducting transcriptomic analysis of developing leaves in the WT and the three mutants we identified differentially expressed genes (DEGs) in the two frameshift mutant lines and found many down-regulated DEGs enriched in photosynthesis, heme, tetrapyrrole binding, and antioxidant activity. In addition, we predicted many target genes of ZmANT1 and chose 13 of them to confirm binding of ZmANT1 to their promoters. Based on the above observations, we proposed a model for ANT1 regulation of cell proliferation and leaf growth, vascular and vein development, chloroplast development, and photosynthesis through its target genes. Our study revealed biological roles of ANT1 in several developmental processes beyond its known roles in plant growth and floral organogenesis.

Plants respond to wounding stress by changing gene expression patterns and inducing the production of hormones including jasmonic acid. This wounding transcriptional response activates specialized metabolism pathways such as the glucosinolate pathways in Arabidopsis thaliana. While the regulatory factors and sequences controlling a subset of wound-response genes are known, it remains unclear how wound response is regulated globally. Here, we how these responses are regulated by incorporating putative cis-regulatory elements, known transcription factor binding sites, in vitro DNA affinity purification sequencing, and DNase I hypersensitive sites to predict genes with different wound-response patterns using machine learning. We observed that regulatory sites and regions of open chromatin differed between genes upregulated at early and late wounding time-points as well as between genes induced by jasmonic acid and those not induced. Expanding on what we currently know, we identified cis-elements that improved model predictions of expression clusters over known binding sites. Using a combination of genome editing, in vitro DNA-binding assays, and transient expression assays using native and mutated cis-regulatory elements, we experimentally validated four of the predicted elements, three of which were not previously known to function in wound-response regulation. Our study provides a global model predictive of wound response and identifies new regulatory sequences important for wounding without requiring prior knowledge of the transcriptional regulators.

Revealing the contributions of genes to plant phenotype is frequently challenging because loss-of-function effects may be subtle or masked by varying degrees of genetic redundancy. Such effects can potentially be detected by measuring plant fitness, which reflects the cumulative effects of genetic changes over the lifetime of a plant. However, fitness is challenging to measure accurately, particularly in species with high fecundity and relatively small propagule sizes such as Arabidopsis thaliana. An image segmentation-based method using the software ImageJ and an object detection-based method using the Faster Region-based Convolutional Neural Network (R-CNN) algorithm were used for measuring two Arabidopsis fitness traits: seed and fruit counts. The segmentation-based method was error-prone (correlation between true and predicted seed counts, r

Iron (Fe) is a key cofactor in many cellular redox processes, including respiration and photosynthesis. Plant Fe deficiency (-Fe) activates a complex regulatory network which coordinates root Fe uptake and distribution to sink tissues, while avoiding over-accumulation of Fe and other metals to toxic levels. In Arabidopsis (Arabidopsis thaliana), FIT (FER-LIKE FE DEFICIENCY-INDUCED TRANSCRIPTION FACTOR), a bHLH transcription factor (TF), is required for up-regulation of root Fe acquisition genes. However, other root and shoot -Fe-induced genes involved in Fe allocation and signaling are FIT-independent. The cis-regulatory code, i.e. the cis-regulatory elements (CREs) and their combinations that regulate plant -Fe-responses, remains largely elusive. Using Arabidopsis genome and transcriptome data, we identified over 100 putative CREs (pCREs) that were predictive of -Fe-induced up-regulation of genes in root tissue. We used large-scale in vitro TF binding data, association with FIT-dependent or FIT-independent co-expression clusters, positional bias, and evolutionary conservation to assess pCRE properties and possible functions. In addition to bHLH and MYB TFs, also B3, NAC, bZIP, and TCP TFs might be important regulators for -Fe responses. Our approach uncovered IDE1 (Iron Deficiency-responsive Element 1), a -Fe response CRE in grass species, to be conserved in regulating genes for biosynthesis of Fe-chelating compounds also in Arabidopsis. Our findings provide a comprehensive source of cis-regulatory information for -Fe-responsive genes, that advances our mechanistic understanding and informs future efforts in engineering plants with more efficient Fe uptake or transport systems.One sentence summary>100 putative cis-regulatory elements robustly predict Arabidopsis root Fe deficiency-responses in computational models, and shed light on the mechanisms of transcriptional regulation.

SummaryRevealing the contributions of genes to plant phenotype is frequently challenging because the effects of loss of gene function may be subtle or be masked by genetic redundancy. Such effects can potentially be detected by measuring plant fitness, which reflects the cumulative effects of genetic changes over the lifetime of a plant. However, fitness is challenging to measure accurately, particularly in species with high fecundity and relatively small propagule sizes such as Arabidopsis thaliana.An image segmentation-based (ImageJ) and a Faster Region Based Convolutional Neural Network (R-CNN) approach were used for measuring two Arabidopsis fitness traits: seed and fruit counts.Although straightforward to use, ImageJ was error-prone (correlation between true and predicted seed counts, r2=0.849) because seeds touching each other were undercounted. In contrast, Faster R-CNN yielded near perfect seed counts (r2=0.9996) and highly accurate fruit counts (r2=0.980).By examining seed counts, we were able to reveal fitness effects for genes that were previously reported to have no or condition-specific loss-of-function phenotypes.Our study provides models to facilitate the investigation of Arabidopsis fitness traits and demonstrates the importance of examining fitness traits in the study of gene functions.

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 posttranslational 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 holdout, 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.

Background Availability of plant genome sequences has led to significant advances. However, with few exceptions, the great majority of existing genome assemblies are derived from short read sequencing technologies with highly uneven read coverages indicative of sequencing and assembly issues that could significantly impact any downstream analysis of plant genomes. In tomato for example, 0.6% (5.1 Mb) and 9.7% (79.6 Mb) of short-read based assembly had significantly higher and lower coverage compared to background, respectively. Results To understand what the causes may be for such uneven coverage, we first established machine learning models capable of predicting genomic regions with variable coverages and found that high coverage regions tend to have higher simple sequence repeat and tandem gene densities compared to background regions. To determine if the high coverage regions were misassembled, we examined a recently available tomato long-read based assembly and found that 27.8% (1.41 Mb) of high coverage regions were potentially misassembled of duplicate sequences, compared to 1.4% in background regions. In addition, using a predictive model that can distinguish correctly and incorrectly assembled high coverage regions, we found that misassembled, high coverage regions tend to be flanked by simple sequence repeats, pseudogenes, and transposon elements. Conclusions Our study provides insights on the causes of variable coverage regions and a quantitative assessment of factors contributing to plant genome misassembly when using short reads and the generality of these causes and factors should be tested further in other species.

Set A and Set B data

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 has not yet been established. In addition, the relative importance of these characteristics for genetic redundancy remains unclear. Here, we establish machine learning models for predicting whether a gene pair is likely redundant or not in the model plant Arabidopsis thaliana. Benchmark gene pairs were classified 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 affected model performance significantly. 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. Predictions were then tested using phenotype data withheld from model building and validated using well-characterized, redundant and nonredundant gene pairs. This genetic redundancy model sheds light on characteristics that may contribute to long-term maintenance of paralogs that are seemingly functionally redundant, and will ultimately allow for more targeted generation of functionally informative double mutants, advancing functional genomic studies.

Plant metabolites produced via diverse pathways are important for plant survival, human nutrition and medicine. However, 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. Utilizing >600 expression values and similarity data combinations from tomato, three strategies for predicting membership in 85 pathways were explored: naive prediction (identifying pathways with the most similarly expressed genes), unsupervised and supervised learning. Optimal predictions for different pathways require distinct data combinations that, in some cases, are indicative of biological processes relevant to pathway functions. Naive prediction produced higher error rates compared with machine learning methods. In 52 pathways, unsupervised learning performed better than a supervised approach, which may be due to the limited availability of training data. Furthermore, using gene-to-pathway expression similarities led to prediction models that outperformed those based simply on gene expression levels. Our study highlights the need to extensively explore expression-based features and prediction strategies to maximize the accuracy of metabolic pathway membership assignment. We anticipate that the prediction framework outlined here can be applied to other species and also be used to improve plant pathway annotation.

Plants respond to wounding stress by changing gene expression patterns and inducing jasmonic acid (JA), as well as other plant hormones. This includes activating some specialized metabolism pathways, including the glucosinolate pathways, in the case of Arabidopsis thaliana. We model how these responses are regulated by using machine learning to incorporate putative cis-regulatory elements (pCREs), known transcription factor binding sites from literature, in-vitro DNA affinity purification sequencing (DAP-seq) and DNase I hypersensitive sites to predict gene expression for genes clustered by their wound response using machine learning. We found temporal patterns where regulatory sites and regions of open chromatin differed between clusters of genes up-regulated at early and late wounding time points as well as clusters where JA response was induced relative to clusters where JA response was not induced. Overall, we identified pCREs that improved model predictions of expression clusters over known binding sites. We discovered 4,255 pCREs related to wound response at different time points and 2,569 pCREs related to differences between JA-induced and non-JA induced wound response. In addition, pCREs found to be important at different wounding time points were mapped to the promoters of genes in a glucosinolate biosynthesis pathway indicating regulation of this pathway under wounding stress. Finally, we experimentally validated a predicted cis-regulatory element, CCGCGT, showing that knock-out via CRISPR-Cas9 reduces gene expression in response to wounding.

Time-series transcriptomes of a biological process obtained under different conditions are useful for identifying the regulators of the process and their regulatory networks. However, such data are 3D (gene expression, time, and condition), and there is currently no method that can deal with their full complexity. Here, we developed a method that avoids time-point alignment and normalization between conditions. We applied it to analyze time-series transcriptomes of developing maize leaves under light–dark cycles and under total darkness and obtained eight time-ordered gene coexpression networks (TO-GCNs), which can be used to predict upstream regulators of any genes in the GCNs. One of the eight TO-GCNs is light-independent and likely includes all genes involved in the development of Kranz anatomy, which is a structure crucial for the high efficiency of photosynthesis in C 4 plants. Using this TO-GCN, we predicted and experimentally validated a regulatory cascade upstream of SHORTROOT1 , a key Kranz anatomy regulator. Moreover, we applied the method to compare transcriptomes from maize and rice leaf segments and identified regulators of maize C 4 enzyme genes and RUBISCO SMALL SUBUNIT2 . Our study provides not only a powerful method but also novel insights into the regulatory networks underlying Kranz anatomy development and C 4 photosynthesis.

Gene co-expression analyses provide a powerful tool to determine gene associations. The interaction of transcription factors (TFs) with their target genes is an essential step in gene regulation, yet to what extent TFs-target gene associations are recovered in co-expression studies remains unclear. Using the wealth of data available for Arabidopsis, we show here that protein-DNA interactions are overall poor indicators of TF-target co-expression, yet the inclusion of TF-TF interaction information significantly enhance co-expression signals. These results highlight the impact of combinatorial gene control on such gene association networks. We integrated this information to predict higher-order regulatory complexes, which are difficult to identify experimentally. We demonstrate that genes strongly co-expressed with a TF are also enriched in indirect targets. Our results have significant implications on the empirical understanding of complex gene regulatory networks and transcription factor function, and the significance of co-expression from the perspective of protein-protein and protein-DNA interactions.

A signaling complex comprising members of the LORELEI (LRE)-LIKE GPI-anchored protein (LLG) and Catharanthus roseus RECEPTOR-LIKE KINASE 1-LIKE (CrRLK1L) families perceive RAPID ALKALINIZATION FACTOR (RALF) peptides and regulate growth, reproduction, immunity, and stress responses in Arabidopsis. Genes encoding these proteins are members of multi-gene families in most angiosperms and could generate thousands of signaling complex variants. However, the link(s) between expansion of these gene families and the functional diversification of this critical signaling complex as well as the evolutionary factors underlying the maintenance of gene duplicates remain unknown. Here, we investigated LLG gene family evolution, function, and expression in angiosperms. We found that LLGs in monocots and eudicots are descendants of a duplication early in angiosperm evolution and that both ancient and recent LLG duplicates are retained. Complementation and expression analysis showed that expression divergence of LLGs (regulatory subfunctionalization), rather than functional divergence, explains the retention of paralogs in Brassicales. All but one extant monocot and eudicot species examined maintained an LLG copy with preferential expression in male reproductive tissues, with the other duplicate copies showed highest levels of expression in female or vegetative tissues. Interestingly, the single LLG copy in Amborella (sister to all other angiosperms) is expressed vastly higher in male compared to female reproductive or vegetative tissues. Reconstruction of expression evolution showed that the highest inferred expression levels for the single copy ancestral angiosperm LLG was in male reproductive tissues. We propose that expression divergence played an important role in maintenance of LLG duplicates in angiosperms.One Sentence SummaryExpression divergence played an important role in maintenance of two sub-groups of LLG duplicates in angiosperms

Plants balance their competing requirements for growth and stress tolerance via a sophisticated regulatory circuitry that controls responses to the external environments. We have identified a plant-specific gene, COST1 ( constitutively stressed 1 ), that is required for normal plant growth but negatively regulates drought resistance by influencing the autophagy pathway. An Arabidopsis thaliana cost1 mutant has decreased growth and increased drought tolerance, together with constitutive autophagy and increased expression of drought-response genes, while overexpression of COST1 confers drought hypersensitivity and reduced autophagy. The COST1 protein is degraded upon plant dehydration, and this degradation is reduced upon treatment with inhibitors of the 26S proteasome or autophagy pathways. The drought resistance of a cost1 mutant is dependent on an active autophagy pathway, but independent of other known drought signaling pathways, indicating that COST1 acts through regulation of autophagy. In addition, COST1 colocalizes to autophagosomes with the autophagosome marker ATG8e and the autophagy adaptor NBR1, and affects the level of ATG8e protein through physical interaction with ATG8e, indicating a pivotal role in direct regulation of autophagy. We propose a model in which COST1 represses autophagy under optimal conditions, thus allowing plant growth. Under drought, COST1 is degraded, enabling activation of autophagy and suppression of growth to enhance drought tolerance. Our research places COST1 as an important regulator controlling the balance between growth and stress responses via the direct regulation of autophagy.

Plants respond to their environment by dynamically modulating gene expression. A powerful approach for understanding how these responses are regulated is to integrate information about cis-regulatory elements (CREs) into models called cis-regulatory codes. Transcriptional response to combined stress is typically not the sum of the responses to the individual stresses. However, cis-regulatory codes underlying combined stress response have not been established. Here we modeled transcriptional response to single and combined heat and drought stress in Arabidopsis thaliana. We grouped genes by their pattern of response (independent, antagonistic and synergistic) and trained machine learning models to predict their response using putative CREs (pCREs) as features (median F-measure = 0.64). We then developed a deep learning approach to integrate additional omics information (sequence conservation, chromatin accessibility and histone modification) into our models, improving performance by 6.2%. While pCREs important for predicting independent and antagonistic responses tended to resemble binding motifs of transcription factors associated with heat and/or drought stress, important synergistic pCREs resembled binding motifs of transcription factors not known to be associated with stress. These findings demonstrate how in silico approaches can improve our understanding of the complex codes regulating response to combined stress and help us identify prime targets for future characterization.

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.

Uncovering genes encoding enzymes responsible for the biosynthesis of diverse plant metabolites is essential for metabolic engineering and production of plant metabolite-derived medicine. With the availability of multi-omics data for an ever-increasing number of plant species and the development of computational approaches, the metabolic pathways of many important plant compounds can be predicted, complementing a more traditional genetic and/or biochemical approach. Here, we summarize recent progress in predicting plant metabolic pathways using genome, transcriptome, proteome, interactome, and/or metabolome data, and the utility of integrating these data with machine learning to further improve metabolic pathway predictions.