Predicting transcriptional responses to cold stress across plant species

Abstract

Although genome sequence assemblies are available for a growing number of plant species, gene expression responses to stimuli have been catalogued for only a subset of these species. Many genes show altered transcription patterns in response to abiotic stresses. However, orthologous genes in related species often exhibit different responses to a given stress. Accordingly, data on the regulation of gene expression in one species are not reliable predictors of orthologous gene responses in a related species. Here, we trained a supervised classification algorithm to identify genes that transcriptionally respond to cold stress. A model trained with only features calculated directly from genome assemblies exhibited only modest decreases in performance relative to models trained using genomic, chromatin, and evolution/diversity features. Models trained with data from one species successfully predicted which genes would respond to cold stress in other related species. Cross-species predictions remained accurate when training was performed in cold-sensitive species and predictions were performed in cold-tolerant species and vice versa. Models trained with data on gene expression in multiple species outperformed models trained with data from any single species. These results suggest that classifiers trained on stress data from well-studied species may suffice for predicting gene expression patterns in related, less-studied species with sequenced genomes.

Machine learning approaches to identify core and dispensable genes in pangenomes

“…Previous applications of machine learning across species yielded similar results (Lee, Karchin, and Beer 2011; Chen, Fish, and Capra 2018; Kelley, n.d.; Mejía-Guerra and Buckler 2019). For example, Meng et al trained machine learning models to identify cold-responsive genes across a few different grass species (Meng et al 2021). Consistent with our results, models performed best when trained and tested in the same species.…”

A review of artificial intelligence-assisted omics techniques in plant defense: current trends and future directions

“…In a study led by Meng et al (2021), supervised classification models were employed to identify genes responding transcriptionally to cold stress. Surprisingly, models trained solely with features derived from genome assemblies displayed modest reductions in performance compared to those incorporating a wider range of data.…”

Temporal regulation of cold transcriptional response in switchgrass

“…The switchgrass cold response RNA-seq data were from a published study of a time course (0.5, 1, 3, 6, 16, and 24 hrs) under cold treatment (6°C) with paired control samples (29°C/ 23°C in a 12-h/12-h day/night cycle) (Meng et al, 2021). Switchgrass transcriptomes under three other stress conditions were from three published studies [Dehydration ( (Zhang et al, 2018)), salt ), and drought ( (Zuo et al, 2018)].…”

“…In switchgrass, coexpression analysis has been used to establish the potential transcriptional regulatory networks in heat, drought, and biotic stress conditions (Pingault et al, 2020;Hayford et al, 2022;Zhou et al, 2022). Recently, a comprehensive, transcriptomic study on several panicoid grasses, including switchgrass, revealed that machine learning approaches can be implemented to predict cold stress responses of genes within and between species based on nucleotide frequencies in promoter regions of genes, among other features (Meng et al, 2021). Beyond nucleotide frequencies, a similar approach using longer nucleotide sequences (i.e., k-mers) can identify putative cisregulatory elements that are regulatory switches of gene expression under cold stress in switchgrass.…”

“…In this study, we aim to apply a similar, machine-learning based approach in switchgrass to assess the involvement of CBFdependent components of cold response regulation and identify other cis-regulatory mechanisms. Using existing cold stress time course transcriptomes of switchgrass (Meng et al, 2021), we first identified temporally cold-responsive genes. To test the extent to which the temporal cold transcriptional response at different cold treatment duration can be explained using potential cisregulatory sequences, we built machine learning models to predict genes that are up-and down-regulated upon cold treatment in the time course experiment using k-mers enriched among up-or down-regulated genes.…”