Identification of new marker genes from plant single‐cell RNA‐seq data using interpretable machine learning methods

Abstract

Summary An essential step in the analysis of single‐cell RNA sequencing data is to classify cells into specific cell types using marker genes. In this study, we have developed a machine learning pipeline called single‐cell predictive marker (SPmarker) to identify novel cell‐type marker genes in the Arabidopsis root. Unlike traditional approaches, our method uses interpretable machine learning models to select marker genes. We have demonstrated that our method can: assign cell types based on cells that were labelled using published methods; project cell types identified by trajectory analysis from one data set to other data sets; and assign cell types based on internal GFP markers. Using SPmarker, we have identified hundreds of new marker genes that were not identified before. As compared to known marker genes, the new marker genes have more orthologous genes identifiable in the corresponding rice single‐cell clusters. The new root hair marker genes also include 172 genes with orthologs expressed in root hair cells in five non‐Arabidopsis species, which expands the number of marker genes for this cell type by 35–154%. Our results represent a new approach to identifying cell‐type marker genes from scRNA‐seq data and pave the way for cross‐species mapping of scRNA‐seq data in plants.

Reference genes for quantitative Arabidopsis single molecule RNA fluorescencein situhybridization

“…To overcome this, we recommend use of multiplexed probe sets to simultaneously label cell-type specific RNA. Machine learning algorithms applied to single cell transcriptome datasets have recently expanded the pool of suitable gene candidates ( Yan et al , 2022 ).…”

Extended discrete gene regulatory network model for theArabidopsis thalianaroot-hair cell fate

“…On the first point, there are several recent approaches that have described the expression profiles of root cells and have tried to establish the developmental trajectories of the different cell types at a single-cell resolution not only for WT but also for some mutants and ambient conditions [19,28,36,79,98,111,147] and some of these approaches even take advantage of machine-learning methods to generate better quality data to identify cell-type marker genes from scRNA-seq data [143]. This approach would allow us to extend our GRN with key regulators that were not incorporated in our model by reconstructing regulatory networks based on massive data [142, 145] and to compare previously proposed conserved loops and incorporate new ones that are able to describe the dynamic behavior of epidermal differentiation in different developmental contexts. These new interactions can be validated in the context of our model to assess their relevance.…”

Cross-species single-cell annotation with orthologous marker gene groups

“…Single-cell RNA sequencing (scRNA-seq) technology has been widely used in identifying and characterizing various cell types in numerous tissues from different plant species [1][2][3][4][5][6] . Efforts to apply this technology in the model plant Arabidopsis have bene ted from the extensive knowledge of cell-type identity markers 7,8 . Contrastingly, accurate labeling of cell types in other plant species remains a challenge due to the scarcity of known cell-type marker genes 9 .…”

“…One way to address this problem is to increase the number of marker genes (i.e., N > 200) with the same statistical method but reduce the speci city threshold of these genes. To increase the number of markers without sacri cing the speci city, we have demonstrated that applying feature selection in machine learning enabled us to identify markers distinct from those identi ed by statistical approaches 7 . Therefore, we employed two additional machine-learning (ML) methods (SHAP-RF and SVM) to identify the top 200 marker genes for each cell-type cluster in each species.…”