AUTHOR=Zhu Xiaoshu , Li Jian , Li Hong-Dong , Xie Miao , Wang Jianxin 

TITLE=Sc-GPE: A Graph Partitioning-Based Cluster Ensemble Method for Single-Cell

JOURNAL=Frontiers in Genetics

VOLUME=Volume 11 - 2020

YEAR=2020

URL=https://www.frontiersin.org/journals/genetics/articles/10.3389/fgene.2020.604790

DOI=10.3389/fgene.2020.604790

ISSN=1664-8021

ABSTRACT=Clustering is an efficient way to analyze single-cell RNA sequencing data. It is commonly used to identify cell types, which can help understanding cell differentiation processes. However, different clustering results will be obtained from different single-cell clustering methods, sometimes including conflicting conclusions, the biologists, will fail to get the right clustering results and interpret the biological significance. The cluster ensemble strategy can be an effective solution for the problem. As the graph partitioning-based clustering methods are good at clustering single-cell, we developed Sc-GPE, a novel cluster ensemble method combining five single-cell graph partitioning-based clustering methods. The five methods are SNN-cliq, PhenoGraph, SC3, SSNN-Louvain, and MPGS-Louvain. In Sc-GPE, a consensus matrix is constructed based on the five clustering solutions, which calculates the probability that the cell pairs are divided into the same cluster. It solves the problem in the hypegraph-based ensemble approach that different cluster labels from the individual clustering method lead it to be difficult to find the correspondence cluster labels across all methods. Then, a weighted consensus matrix is constructed to distinguish the different importance of each method in clustering by scoring the importance of the individual clustering methods. Finally, hierarchical clustering is performed on the weighted consensus matrix to cluster cells. To evaluate the performance, we compare Sc-GPE with the individual clustering methods and the state-of-the-art SAME-clustering on twelve single-cell RNA-seq datasets. The results show that Sc-GPE obtains the best average performance, and achieves the highest NMI and ARI value in five datasets.