Clustering of Single-Cell Transcriptome Data Based on Evolutionary Algorithm in Assimilation with Fuzzy C-Means

Abstract

Single-cell transcriptome sequencing (scRNA-seq) technology enables to analyze the RNA expression of each cell over a different instance of time. This provides the path to identify different patterns of gene expression through gene clustering. To analyze the gene expression pattern of scRNA-seq data, soft clustering plays a crucial role. This can exploit the underlying structure of an unknown set of patterns. To assign data points to different clusters, a newly developed weighted distance measure is utilized in the paper. Here, minimization of variance within the cluster is posed as the objective function. To obtain a good partition of the data, different search operator of gray wolf optimizer (GWO) technique is explored in conjunction with fuzzy C-means (FCM) algorithm. The proposed (fuzzy C-means and GWO) is then compared with some recent scRNA-seq clustering techniques such as Seurat, SC3, CIDR, t-SNE followed by k-means and ensemble clustering using cluster validity index. Analysis of the clustering result demonstrated that the application of an evolutionary-based optimization technique jointly with the FCM algorithm can cluster and identify the transcriptomics gene effectively.