Research on Preprocessing and Downstream Analysis of Single-Cell Transcriptome Data Based on Seurat and Harmony

Abstract

To systematically decipher the biological information embedded in single-cell transcriptomic data, this study utilized the 10X Genomics PBMC 3k dataset as the research subject and integrated the Seurat package in R with the Harmony algorithm to conduct a comprehensive analytical workflow—from raw count matrices to cell annotation and differential expression analysis. The study first applied quality filtering at both gene and cell levels to remove low-quality data, followed by effective batch effect correction using the Harmony algorithm. Subsequent steps included linear dimensionality reduction via PCA, non-linear dimensionality reduction through UMAP, and cell clustering based on Shared Nearest Neighbor (SNN) graph partitioning. Cell type identification was accomplished by combining automated annotation with SingleR and manual validation via marker genes, ultimately leading to the screening of differentially expressed genes across distinct cell subpopulations. Results demonstrated that after quality filtering, 4,200 high-quality cells and 25,000 valid genes were retained; the Harmony algorithm completely eliminated batch biases between stimulated and control groups. Clustering at a resolution of 0.8 yielded 11 biologically meaningful PBMC subpopulations, successfully identifying major immune cell types—such as T cells, B cells, and monocytes—along with their characteristic marker genes. The established single-cell transcriptomic data analysis pipeline provides a robust methodological framework for resolving cellular heterogeneity and characterizing feature genes of cell subtypes, thereby laying a data foundation for functional studies of immune cells.