Leiden Clustering Based on Single-cell Sequencing Data of Human Bone Marrow

Abstract

As single-cell sequencing technology has gradually become a popular method for obtaining effective data in biology, the experimental steps corresponding to different experimental goals are also very different, which often causes researchers to make mistakes in choosing data processing methods. In response to this situation, this paper selects a part of the NeurIPS 2021 benchmark dataset of openproblem because the data processing difficulty is moderate, the data noise is low, and the dataset is bone marrow mononuclear cell data of healthy human donors, which is representative in cell processing, and conducts an in-depth compilation and analysis of the dataset, summarizing a set of more universal single-cell sequencing experimental steps. This paper first filters cells and genes by setting different thresholds of corresponding indicators to achieve the purpose of data preprocessing; in terms of data dimensionality reduction, this paper uses principal component analysis (PCA) to reduce the dimensionality of the data and mark more than 2,000 highly variable genes. Then clustering the cells, and this paper divides the cell clusters into 27 categories through the Leiden clustering method. Meanwhile, this paper also identifies and analyzes the marker genes based on the cell clusters obtained by clustering. Through the research of this article, it is found that compared with traditional sequencing methods (such as K-means, hierarchical clustering, Louvain clustering, etc.), Leiden clustering is the most locally distributed for all subsets of all communities. Salient features, and considering that Louvain clustering’s lack of community connectivity is difficult to solve, make Leiden clustering more superior to Louvain clustering in processing single-cell data. At the same time, the research of this paper not only innovates in the method of single-cell sequencing, but also achieves a driving significance in the derivation of its biological functions.