Assessing the evolution of research topics in a biological field using plant science as an example.

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.

High-throughput measurement of plant fitness traits with an object detection method using Faster R-CNN.

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, r2 = 0.849) because seeds touching each other were undercounted. By contrast, the object detection-based algorithm yielded near perfect seed counts (r2 = 0.9996) and highly accurate fruit counts (r2 = 0.980). Comparing seed counts for wild-type and 12 mutant lines revealed fitness effects for three genes; fruit counts revealed the same effects for two genes. Our study provides analysis pipelines and models to facilitate the investigation of Arabidopsis fitness traits and demonstrates the importance of examining fitness traits when studying gene functions.

Predictive Models of Genetic Redundancy in Arabidopsis thaliana.

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.

Optimising the use of gene expression data to predict plant metabolic pathway memberships.

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.

Within- and cross-species predictions of plant specialized metabolism genes using transfer learning.

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.

Robust predictions of specialized metabolism genes through machine learning.

Plant specialized metabolism (SM) enzymes produce lineage-specific metabolites with important ecological, evolutionary, and biotechnological implications. Using Arabidopsis thaliana as 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 1,220 A. thaliana genes 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.

Factors Influencing Gene Family Size Variation Among Related Species in a Plant Family, Solanaceae.

Gene duplication and loss contribute to gene content differences as well as phenotypic divergence across species. However, the extent to which gene content varies among closely related plant species and the factors responsible for such variation remain unclear. Here, using the Solanaceae family as a model and Pfam domain families as a proxy for gene families, we investigated variation in gene family sizes across species and the likely factors contributing to the variation. We found that genes in highly variable families have high turnover rates and tend to be involved in processes that have diverged between Solanaceae species, whereas genes in low-variability families tend to have housekeeping roles. In addition, genes in high- and low-variability gene families tend to be duplicated by tandem and whole genome duplication, respectively. This finding together with the observation that genes duplicated by different mechanisms experience different selection pressures suggest that duplication mechanism impacts gene family turnover. We explored using pseudogene number as a proxy for gene loss but discovered that a substantial number of pseudogenes are actually products of pseudogene duplication, contrary to the expectation that most plant pseudogenes are remnants of once-functional duplicates. Our findings reveal complex relationships between variation in gene family size, gene functions, duplication mechanism, and evolutionary rate. The patterns of lineage-specific gene family expansion within the Solanaceae provide the foundation for a better understanding of the genetic basis underlying phenotypic diversity in this economically important family.

Diversity, expansion, and evolutionary novelty of plant DNA-binding transcription factor families.

Plant transcription factors (TFs) that interact with specific sequences via DNA-binding domains are crucial for regulating transcriptional initiation and are fundamental to plant development and environmental response. In addition, expansion of TF families has allowed functional divergence of duplicate copies, which has contributed to novel, and in some cases adaptive, traits in plants. Thus, TFs are central to the generation of the diverse plant species that we see today. Major plant agronomic traits, including those relevant to domestication, have also frequently arisen through changes in TF coding sequence or expression patterns. Here our goal is to provide an overview of plant TF evolution by first comparing the diversity of DNA-binding domains and the sizes of these domain families in plants and other eukaryotes. Because TFs are among the most highly expanded gene families in plants, the birth and death process of TFs as well as the mechanisms contributing to their retention are discussed. We also provide recent examples of how TFs have contributed to novel traits that are important in plant evolution and in agriculture.This article is part of a Special Issue entitled: Plant Gene Regulatory Mechanisms and Networks, edited by Dr. Erich Grotewold and Dr. Nathan Springer.

Utility and Limitations of Using Gene Expression Data to Identify Functional Associations.

Gene co-expression has been widely used to hypothesize gene function through guilt-by association. However, it is not clear to what degree co-expression is informative, whether it can be applied to genes involved in different biological processes, and how the type of dataset impacts inferences about gene functions. Here our goal is to assess the utility and limitations of using co-expression as a criterion to recover functional associations between genes. By determining the percentage of gene pairs in a metabolic pathway with significant expression correlation, we found that many genes in the same pathway do not have similar transcript profiles and the choice of dataset, annotation quality, gene function, expression similarity measure, and clustering approach significantly impacts the ability to recover functional associations between genes using Arabidopsis thaliana as an example. Some datasets are more informative in capturing coordinated expression profiles and larger data sets are not always better. In addition, to recover the maximum number of known pathways and identify candidate genes with similar functions, it is important to explore rather exhaustively multiple dataset combinations, similarity measures, clustering algorithms and parameters. Finally, we validated the biological relevance of co-expression cluster memberships with an independent phenomics dataset and found that genes that consistently cluster with leucine degradation genes tend to have similar leucine levels in mutants. This study provides a framework for obtaining gene functional associations by maximizing the information that can be obtained from gene expression datasets.

Molecular Evidence for Functional Divergence and Decay of a Transcription Factor Derived from Whole-Genome Duplication in Arabidopsis thaliana.

Functional divergence between duplicate transcription factors (TFs) has been linked to critical events in the evolution of land plants and can result from changes in patterns of expression, binding site divergence, and/or interactions with other proteins. Although plant TFs tend to be retained post polyploidization, many are lost within tens to hundreds of million years. Thus, it can be hypothesized that some TFs in plant genomes are in the process of becoming pseudogenes. Here, we use a pair of salt tolerance-conferring transcription factors, DWARF AND DELAYED FLOWERING1 (DDF1) and DDF2, that duplicated through paleopolyploidy 50 to 65 million years ago, as examples to illustrate potential mechanisms leading to duplicate retention and loss. We found that the expression patterns of Arabidopsis thaliana (At)DDF1 and AtDDF2 have diverged in a highly asymmetric manner, and AtDDF2 has lost most inferred ancestral stress responses. Consistent with promoter disablement, the AtDDF2 promoter has fewer predicted cis-elements and a methylated repetitive element. Through comparisons of AtDDF1, AtDDF2, and their Arabidopsis lyrata orthologs, we identified significant differences in binding affinities and binding site preference. In particular, an AtDDF2-specific substitution within the DNA-binding domain significantly reduces binding affinity. Cross-species analyses indicate that both AtDDF1 and AtDDF2 are under selective constraint, but among A. thaliana accessions, AtDDF2 has a higher level of nonsynonymous nucleotide diversity compared with AtDDF1. This may be the result of selection in different environments or may point toward the possibility of ongoing functional decay despite retention for millions of years after gene duplication.

Characteristics and significance of intergenic polyadenylated RNA transcription in Arabidopsis.

The Arabidopsis (Arabidopsis thaliana) genome is the most well-annotated plant genome. However, transcriptome sequencing in Arabidopsis continues to suggest the presence of polyadenylated (polyA) transcripts originating from presumed intergenic regions. It is not clear whether these transcripts represent novel noncoding or protein-coding genes. To understand the nature of intergenic polyA transcription, we first assessed its abundance using multiple messenger RNA sequencing data sets. We found 6,545 intergenic transcribed fragments (ITFs) occupying 3.6% of Arabidopsis intergenic space. In contrast to transcribed fragments that map to protein-coding and RNA genes, most ITFs are significantly shorter, are expressed at significantly lower levels, and tend to be more data set specific. A surprisingly large number of ITFs (32.1%) may be protein coding based on evidence of translation. However, our results indicate that these "translated" ITFs tend to be close to and are likely associated with known genes. To investigate if ITFs are under selection and are functional, we assessed ITF conservation through cross-species as well as within-species comparisons. Our analysis reveals that 237 ITFs, including 49 with translation evidence, are under strong selective constraint and relatively distant from annotated features. These ITFs are likely parts of novel genes. However, the selective pressure imposed on most ITFs is similar to that of randomly selected, untranscribed intergenic sequences. Our findings indicate that despite the prevalence of ITFs, apart from the possibility of genomic contamination, many may be background or noisy transcripts derived from "junk" DNA, whose production may be inherent to the process of transcription and which, on rare occasions, may act as catalysts for the creation of novel genes.

Diversity, classification and function of the plant protein kinase superfamily.

Eukaryotic protein kinases belong to a large superfamily with hundreds to thousands of copies and are components of essentially all cellular functions. The goals of this study are to classify protein kinases from 25 plant species and to assess their evolutionary history in conjunction with consideration of their molecular functions. The protein kinase superfamily has expanded in the flowering plant lineage, in part through recent duplications. As a result, the flowering plant protein kinase repertoire, or kinome, is in general significantly larger than other eukaryotes, ranging in size from 600 to 2500 members. This large variation in kinome size is mainly due to the expansion and contraction of a few families, particularly the receptor-like kinase/Pelle family. A number of protein kinases reside in highly conserved, low copy number families and often play broadly conserved regulatory roles in metabolism and cell division, although functions of plant homologues have often diverged from their metazoan counterparts. Members of expanded plant kinase families often have roles in plant-specific processes and some may have contributed to adaptive evolution. Nonetheless, non-adaptive explanations, such as kinase duplicate subfunctionalization and insufficient time for pseudogenization, may also contribute to the large number of seemingly functional protein kinases in plants.

Evolution of stress-regulated gene expression in duplicate genes of Arabidopsis thaliana.

Due to the selection pressure imposed by highly variable environmental conditions, stress sensing and regulatory response mechanisms in plants are expected to evolve rapidly. One potential source of innovation in plant stress response mechanisms is gene duplication. In this study, we examined the evolution of stress-regulated gene expression among duplicated genes in the model plant Arabidopsis thaliana. Key to this analysis was reconstructing the putative ancestral stress regulation pattern. By comparing the expression patterns of duplicated genes with the patterns of their ancestors, duplicated genes likely lost and gained stress responses at a rapid rate initially, but the rate is close to zero when the synonymous substitution rate (a proxy for time) is > approximately 0.8. When considering duplicated gene pairs, we found that partitioning of putative ancestral stress responses occurred more frequently compared to cases of parallel retention and loss. Furthermore, the pattern of stress response partitioning was extremely asymmetric. An analysis of putative cis-acting DNA regulatory elements in the promoters of the duplicated stress-regulated genes indicated that the asymmetric partitioning of ancestral stress responses are likely due, at least in part, to differential loss of DNA regulatory elements; the duplicated genes losing most of their stress responses were those that had lost more of the putative cis-acting elements. Finally, duplicate genes that lost most or all of the ancestral responses are more likely to have gained responses to other stresses. Therefore, the retention of duplicates that inherit few or no functions seems to be coupled to neofunctionalization. Taken together, our findings provide new insight into the patterns of evolutionary changes in gene stress responses after duplication and lay the foundation for testing the adaptive significance of stress regulatory changes under highly variable biotic and abiotic environments.

Evolutionary and expression signatures of pseudogenes in Arabidopsis and rice.

Pseudogenes (Psi) are nonfunctional genomic sequences resembling functional genes. Knowledge of Psis can improve genome annotation and our understanding of genome evolution. However, there has been relatively little systemic study of Psis in plants. In this study, we characterized the evolution and expression patterns of Psis in Arabidopsis (Arabidopsis thaliana) and rice (Oryza sativa). In contrast to animal Psis, many plant Psis experienced much stronger purifying selection. In addition, plant Psis experiencing stronger selective constraints tend to be derived from relatively ancient duplicates, suggesting that they were functional for a relatively long time but became Psis recently. Interestingly, the regions 5' to the first stops in the Psis have experienced stronger selective constraints compared with 3' regions, suggesting that the 5' regions were functional for a longer period of time after the premature stops appeared. We found that few Psis have expression evidence, and their expression levels tend to be lower compared with annotated genes. Furthermore, Psis with expressed sequence tags tend to be derived from relatively recent duplication events, indicating that Psi expression may be due to insufficient time for complete degeneration of regulatory signals. Finally, larger protein domain families have significantly more Psis in general. However, while families involved in environmental stress responses have a significant excess of Psis, transcription factors and receptor-like kinases have lower than expected numbers of Psis, consistent with their elevated retention rate in plant genomes. Our findings illustrate peculiar properties of plant Psis, providing additional insight into the evolution of duplicate genes and benefiting future genome annotation.

Evolutionary history and stress regulation of plant receptor-like kinase/pelle genes.

Receptor-Like Kinase (RLK)/Pelle genes play roles ranging from growth regulation to defense response, and the dramatic expansion of this family has been postulated to be crucial for plant-specific adaptations. Despite this, little is known about the history of or the factors that contributed to the dramatic expansion of this gene family. In this study, we show that expansion coincided with the establishment of land plants and that RLK/Pelle subfamilies were established early in land plant evolution. The RLK/Pelle family expanded at a significantly higher rate than other kinases, due in large part to expansion of a few subfamilies by tandem duplication. Interestingly, these subfamilies tend to have members with known roles in defense response, suggesting that their rapid expansion was likely a consequence of adaptation to fast-evolving pathogens. Arabidopsis (Arabidopsis thaliana) expression data support the importance of RLK/Pelles in biotic stress response. We found that hundreds of RLK/Pelles are up-regulated by biotic stress. Furthermore, stress responsiveness is correlated with the degree of tandem duplication in RLK/Pelle subfamilies. Our findings suggest a link between stress response and tandem duplication and provide an explanation for why a large proportion of the RLK/Pelle gene family is found in tandem repeats. In addition, our findings provide a useful framework for potentially predicting RLK/Pelle stress functions based on knowledge of expansion pattern and duplication mechanism. Finally, we propose that the detection of highly variable molecular patterns associated with specific pathogens/parasites is the main reason for the up-regulation of hundreds of RLK/Pelles under biotic stress.

Importance of lineage-specific expansion of plant tandem duplicates in the adaptive response to environmental stimuli.

Plants have substantially higher gene duplication rates compared with most other eukaryotes. These plant gene duplicates are mostly derived from whole genome and/or tandem duplications. Earlier studies have shown that a large number of duplicate genes are retained over a long evolutionary time, and there is a clear functional bias in retention. However, the influence of duplication mechanism, particularly tandem duplication, on duplicate retention has not been thoroughly investigated. We have defined orthologous groups (OGs) between Arabidopsis (Arabidopsis thaliana) and three other land plants to examine the functional bias of retained duplicate genes during vascular plant evolution. Based on analysis of Gene Ontology categories, it is clear that genes in OGs that expanded via tandem duplication tend to be involved in responses to environmental stimuli, while those that expanded via nontandem mechanisms tend to have intracellular regulatory roles. Using Arabidopsis stress expression data, we further demonstrated that tandem duplicates in expanded OGs are significantly enriched in genes that are up-regulated by biotic stress conditions. In addition, tandem duplication of genes in an OG tends to be highly asymmetric. That is, expansion of OGs with tandem genes in one organismal lineage tends to be coupled with losses in the other. This is consistent with the notion that these tandem genes have experienced lineage-specific selection. In contrast, OGs with genes duplicated via nontandem mechanisms tend to experience convergent expansion, in which similar numbers of genes are gained in parallel. Our study demonstrates that the expansion of gene families and the retention of duplicates in plants exhibit substantial functional biases that are strongly influenced by the mechanism of duplication. In particular, genes involved in stress responses have an elevated probability of retention in a single-lineage fashion following tandem duplication, suggesting that these tandem duplicates are likely important for adaptive evolution to rapidly changing environments.

The MADS domain factors AGL15 and AGL18 act redundantly as repressors of the floral transition in Arabidopsis.

The developmental roles of AGL15 and AGL18, members of the AGL15-like clade of MADS domain regulatory factors, have not been defined previously. Analysis of transgenic Arabidopsis plants showed that overexpression of AGL18 produces the same phenotypic changes as overexpression of AGL15, and the two genes have partially overlapping expression patterns. Functional redundancy was confirmed through analysis of loss-of-function mutants. agl15 agl18 double mutants, but not single mutants, flower early under non-inductive conditions, indicating that AGL15 and AGL18 act in a redundant fashion as repressors of the floral transition. Further genetic analyses and expression studies were used to examine the relationship between AGL15 and AGL18 activity and other regulators of the floral transition. AGL15 and AGL18 act upstream of the floral integrator FT, and a combination of agl15 and agl18 mutations partially suppresses defects in the photoperiod pathway. agl15 agl18 mutations show an additive relationship with mutations in genes encoding other MADS domain floral repressors, and further acceleration of flowering is seen in triple and quadruple mutants under both inductive and non-inductive conditions. Thus, flowering time is determined by the additive effect of multiple MADS domain floral repressors, with important contributions from AGL15 and AGL18.

Identification of quantitative trait loci that regulate Arabidopsis root system size and plasticity.

Root system size (RSS) is a complex trait that is greatly influenced by environmental cues. Hence, both intrinsic developmental pathways and environmental-response pathways contribute to RSS. To assess the natural variation in both types of pathways, we examined the root systems of the closely related Arabidopsis ecotypes Landsberg erecta (Ler) and Columbia (Col) grown under mild osmotic stress conditions. We found that Ler initiates more lateral root primordia, produces lateral roots from a higher percentage of these primordia, and has an overall larger root system than Col under these conditions. Furthermore, although each of these parameters is reduced by osmotic stress in both ecotypes, Ler shows a decreased sensitivity to osmotica. To understand the genetic basis for these differences, QTL for RSS under mild osmotic stress were mapped in a Ler x Col recombinant inbred population. Two robust quantitative trait loci (QTL) were identified and confirmed in near-isogenic lines (NILs). The NILs also allowed us to define distinct physiological roles for the gene(s) at each locus. This study provides insight into the genetic and physiological complexity that determines RSS and begins to dissect the molecular basis for naturally occurring differences in morphology and developmental plasticity in the root system.

Expression of MADS-box genes during the embryonic phase in Arabidopsis.

MADS domain factors play important roles as developmental regulators in plants. In Arabidopsis thaliana, MADS domain proteins have been shown to regulate various processes during the vegetative and reproductive phases. Relatively little is known, however, about family members expressed during the embryonic phase and their function. To determine which MADS-box genes are expressed during the embryonic phase in Arabidopsis, a family-wide survey involving gene-specific primers and RT-PCR was conducted. Transcripts corresponding to 64 (out of 109 total) family members could be detected in RNA samples isolated from embryonic culture tissue. Eight MADS-box genes that appear to be expressed at higher levels during the embryonic phase than in seedlings or in inflorescence apices were identified. The spatial pattern of expression in developing seeds was characterized for four MADS-box genes (FLOWERING LOCUS C, FLOWERING LOCUS M, AGAMOUS-LIKE 15, and AGAMOUS-LIKE 18) using reporter constructs encoding translational fusions to GUS. All four are expressed in cells throughout the endosperm and embryo. Finally, to test the hypothesis that AGAMOUS-LIKE15 (AGL15) and AGAMOUS-LIKE18 (AGL18) play essential roles during the embryonic phase, plants carrying T-DNA insertions that disrupt these genes were isolated. No embryo defects were observed in agl15 or agl18 single mutants or in agl15agl18 double mutants. These results indicate that multiple regulatory pathways that involve MADS domain factors are likely to operate in embryonic tissues, and that genetic and/or functional redundancy are likely to be as prevalent as in other phases of the life cycle.