Seurat analysis
For basic data visualization, a log-transformed gene matrix was imported into R Studio and analyzed by the Seurat package (http://www.satijalab.org/seurat). Cells with fewer than 500 detected genes and greater than 10% mitochondrial gene expression were removed. Moreover, we eliminated genes detected (UMI count >0) in less than three cells and deleted mitochondrial and ribosomal genes. The remaining genes were reduced in dimension by principle component analysis. After quality control and normalization, we used T-distributed Stochastic Neighbor Embedding (t-SNE) to visualize the characteristics of the continuity of subsets. Using other functions of Seurat, including FindVariableFeatures, FindAllMarkers, and Doheatmap, every cluster of collected cells was user-redefined for further analysis.