Consensus-based clustering of single cells by reconstructing cell-to-cell dissimilarity

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) enables the study of cellular heterogeneity in complex tissues.
• Clustering algorithms are crucial for identifying distinct cell types from scRNA-seq data.
• The choice of distance metric significantly impacts the performance of clustering algorithms.

Purpose of the Study:

• To address the limitations of existing distance measures in scRNA-seq data clustering.
• To introduce a novel single-cell clustering method, SD-h, that generates an optimized distance metric.
• To improve the accuracy of cell-type identification in scRNA-seq analyses.

Main Methods:

• Development of the SD-h clustering method, which synthesizes multiple distance measures.
• Application of hierarchical clustering using the newly generated distance measure.
• Validation of SD-h on nine diverse scRNA-seq datasets.

Main Results:

• SD-h demonstrated superior performance compared to existing leading single-cell clustering algorithms.
• The method's ability to generate an applicable distance measure across different datasets was confirmed.
• SD-h achieved more accurate cell-type clustering.

Conclusions:

• The SD-h method offers a robust and accurate approach for cell-type clustering in scRNA-seq data.
• Optimizing distance measures is critical for enhancing the performance of single-cell data analysis.
• SD-h provides a valuable tool for researchers studying cellular heterogeneity.