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

Abstract

An essential step of single-cell RNA sequencing analysis is to classify specific cell types with marker genes in order to dissect the biological functions of each individual cell. In this study, we integrated five published scRNA-seq datasets from the Arabidopsis root containing over 25,000 cells and 17 cell clusters. We have compared the performance of seven machine learning methods in classifying these cell types, and determined that the random forest and support vector machine methods performed best. Using feature selection with these two methods and a correlation method, we have identified 600 new marker genes for 10 root cell types, and more than 70% of these machine learning-derived marker genes were not identified before. We found that these new markers not only can assign cell types consistently as the previously known cell markers, but also performed better than existing markers in several evaluation metrics including accuracy and sensitivity. Markers derived by the random forest method, in particular, were expressed in 89-98% of cells in endodermis, trichoblast, and cortex clusters, which is a 29-67% improvement over known markers. Finally, we have found 111 new orthologous marker genes for the trichoblast in five plant species, which expands the number of marker genes by 58-170% in non-Arabidopsis plants. Our results represent a new approach to identify cell-type marker genes from scRNA-seq data and pave the way for cross-species mapping of scRNA-seq data in plants.

Single-cell genomics in plants: current state, future directions, and hurdles to overcome

“…Nevertheless, other aspects of gene regulation such as transcription factors and their binding motifs are conserved across plants. The potential for classifying cell types in less-well characterized plants on the basis of existing single-cell data from Arabidopsis is well illustrated by a recent study that tested several of the 111 identified marker genes for trichoblast cells in five other plant species ( Yan et al, 2020 ). As the efficacy of machine learning relies critically on data volume, the results will only improve as more and more datasets become available.…”

A Novel Algorithm for Feature Selection Using Penalized Regression with Applications to Single-Cell RNA Sequencing Data

“…Therefore, RFE is not an ideal choice for feature selection in scRNA-seq data. Random forest and its application on scRNA-seq data have been examined thoroughly in other studies [ 8 , 9 , 10 , 11 ]. Moreover, some penalized regression methods are worth exploring to compare their performances for feature selection in scRNA-seq data because these methods were developed primarily for tackling the challenges of “large-p-small-n” problems [ 12 ].…”