Single-Cell RNA-Seq Debiased Clustering via Batch Effect Disentanglement.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables cellular phenotype discovery.
• Batch effects from experimental variations can bias scRNA-seq data, confounding biological insights.
• Existing methods struggle to integrate batch correction and clustering effectively.

Purpose of the Study:

• To develop a robust method for scRNA-seq data clustering that mitigates batch effect bias.
• To disentangle biological variation from non-biological (batch) effects during data partitioning.
• To improve the accuracy and reliability of cell type identification in scRNA-seq datasets.

Main Methods:

• Proposing single-cell RNA-seq debiased clustering (SCDC), an end-to-end clustering approach.
• Implementing a strategy to disentangle biological and non-biological information during data partitioning.
• Evaluating SCDC performance against state-of-the-art clustering and batch integration methods.

Main Results:

• SCDC demonstrates superior qualitative and quantitative performance in six analyses.
• The method effectively handles scRNA-seq data confounded by batch effects.
• SCDC exhibits scalable performance with linearly increasing running time and fixed GPU memory consumption.

Conclusions:

• SCDC offers a powerful solution for debiased clustering of scRNA-seq data.
• The method successfully addresses the challenge of batch effects in single-cell data analysis.
• SCDC is scalable for large datasets, with code availability on Github.