SCEC: A Novel Single-Cell Classification Method Based on Cell-Pair Ensemble Learning

Abstract

With the development of single-cell sequencing technology, the increasing amount of single-cell transcription data enables a deeper understanding of cellular heterogeneity and mechanisms at the cell level. Single-cell type identification is a crucial step of single-cell RNA sequencing (scRNA-seq) data analysis. The classification of single-cells and the discovery of new cell types are of great significance for oncology, immunology, and developmental biology research. Various data preprocessing approaches and classification algorithms have been applied to scRNA-seq data analysis. However, single-cell type identification methods that integrate multiple dimensionality reduction methods and unsupervised classification are still rare. Here, we proposed SCEC (Single Cell Ensemble Classify), an ensemble learning based approach, to integrate the clustering results of different clustering methods and SCEC significantly improves type identification on several representative datasets. Furthermore, the consensus result given by SCEC also shows obvious advantages in the process of data visualization. The advantages in classification and visualization performance make SCEC a promising method for large-scale scRNA-seq data.