scBKAP: A Clustering Model for Single-Cell RNA-Seq Data Based on Bisecting K-Means.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables high-resolution analysis of cellular transcription.
• Existing scRNA-seq clustering methods face challenges with high data dropout rates and the curse of dimensionality.

Purpose of the Study:

• To introduce scBKAP, a novel computational pipeline for robust scRNA-seq data clustering.
• To overcome limitations of existing methods in handling sparse and high-dimensional scRNA-seq data.

Main Methods:

• scBKAP employs an autoencoder network for gene expression reconstruction to mitigate dropout effects.
• A dimensionality reduction model, MPDR (M3Drop and PHATE), is used on reconstructed data.
• Bisecting K-means clustering is applied to the dimensionality-reduced data for cell cluster identification.

Main Results:

• scBKAP demonstrated superior performance compared to nine state-of-the-art methods.
• The pipeline was validated on 21 public scRNA-seq datasets and simulated data.
• Effective alleviation of dropout issues and improved clustering accuracy were observed.

Conclusions:

• scBKAP offers a robust and effective solution for scRNA-seq data clustering.
• The method enhances biological discovery by improving cell type identification from complex datasets.
• The pipeline provides an open-source tool for the research community.