Clustering Deviation Index (CDI): a robust and accurate internal measure for evaluating scRNA-seq data clustering

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) is crucial for understanding cellular heterogeneity.
• Accurate cell clustering is fundamental for reliable downstream scRNA-seq analysis.
• Existing tools for assessing clustering accuracy are limited.

Purpose of the Study:

• To develop a novel metric for evaluating scRNA-seq clustering accuracy.
• To introduce the Clustering Deviation Index (CDI) for assessing clustering performance.
• To demonstrate CDI's utility in optimizing clustering parameters and determining cluster numbers.

Main Methods:

• Developed the Clustering Deviation Index (CDI) to quantify clustering accuracy.
• Applied CDI to both in silico and experimental scRNA-seq datasets.
• Utilized CDI to guide the selection of optimal clustering label sets and parameters.

Main Results:

• CDI effectively measures the deviation of clustering results from scRNA-seq data.
• CDI successfully identified the optimal clustering label set across different datasets.
• CDI informed the selection of optimal tuning parameters and the correct number of clusters.

Conclusions:

• The Clustering Deviation Index (CDI) is a valuable tool for scRNA-seq data analysis.
• CDI enhances the reliability of cell clustering and downstream analyses.
• CDI facilitates the optimization of clustering methods and parameter selection in scRNA-seq studies.