ClusterMatch aligns single-cell RNA-sequencing data at the multi-scale cluster level via stable matching.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables cell type discovery but faces challenges with multi-scale clustering resolutions due to high-dimensional, noisy data.
• Variability in scRNA-seq data complicates accurate identification and characterization of cell populations.

Purpose of the Study:

• To introduce ClusterMatch, a novel stable matching optimization model for aligning scRNA-seq data at the cluster level.
• To address challenges in multi-scale clustering resolutions and data variability in scRNA-seq analysis.
• To improve cell type characterization, data integration, and alignment across different biological conditions.

Main Methods:

• ClusterMatch employs canonical correlation analysis and multi-scale Louvain clustering to identify optimal cluster resolutions.
• A stable matching framework is utilized to align scRNA-seq data in a latent space, preserving interpretability through shared marker genes.
• The model integrates global and local data information for robust analysis.

Main Results:

• ClusterMatch demonstrates efficacy in scRNA-seq data integration, cell type annotation, and cross-species/timepoint alignment.
• The method successfully identifies clusters with optimized resolutions by leveraging both global and local data features.
• Interpretability is maintained by utilizing marker gene sets within the alignment process.

Conclusions:

• ClusterMatch provides a robust framework for scRNA-seq data alignment and cell type discovery.
• The model's ability to handle multi-scale resolutions and data variability enhances its applicability in biological and clinical research.
• ClusterMatch facilitates accurate cell type annotation and data integration, offering a valuable tool for researchers.