Research On Deep Clustering Algorithm For Single Cell RNA Sequencing Data

The development of single-cell RNA sequencing technology allows us to analyze transcriptome data at the single-cell level.With single-cell RNA sequencing,we can get thousands of gene expression values for each cell in a tissue,but we can't get information about the cell types contained in the tissue.We can identify cell types by clustering similar cells together and separating dissimilar cells through cell clustering algorithms.Cell clustering plays an important role in biology and medicine.In this paper,we propose a deep clustering model based on graph convolutional neural network for cell clustering,which combines the strong ability of autoencoder neural networks to learn potential representations and the ability of graph convolutional neural networks to learn higher-order structural relationships of data.In this paper,the coding layer of the trained autoencoder neural network is embedded into the graph convolutional neural network through the transfer operator,so that the information of each layer of the graph convolutional network contains the potential information of the data and the higher-order structural information of the data.At the same time,this paper proposes a dual self-supervised model to train two different neural network modules end-to-end until they converge.Finally,we take the output of the graph convolutional neural network which has learned the potential information of the data and the high-order structure information of the data as the final clustering result.Clustering results of four real datasets show that the model in this paper has a relatively accurate clustering of cells in the datasets,and has more advantages than several existing single-cell clustering methods,which indicates the effectiveness of the model in this paper.