Systematic clustering alignment and feature characterization for single-cell omics using ACE-OF-Clust.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Clustering is crucial for cell type identification in transcriptomic data like scRNA-seq and ST.
• Mixed-membership clustering captures continuous variation but faces challenges integrating results.
• The 'clustering alignment problem' complicates interpretation due to label switching and differing model settings.

Purpose of the Study:

• To introduce ACE-OF-Clust, a novel workflow for single-cell clustering.
• To enable direct comparison and alignment of diverse clustering solutions.
• To improve the interpretability, flexibility, and robustness of single-cell data analysis.

Main Methods:

• ACE-OF-Clust employs a four-step workflow: multiple clustering, alignment, model comparison, and feature identification.
• It directly compares clustering solutions and assesses consistency with annotations.
• Feature-level clustering profiles are used to identify discriminating genes.

Main Results:

• Demonstrated utility on PBMC scRNA-seq, breast cancer ST, and multi-omic single-cell data.
• Quantified cross-omic clustering variability and identified potential cross-omic regulatory links.
• ACE-OF-Clust successfully prioritized genes that distinguish cell types.

Conclusions:

• ACE-OF-Clust provides a scalable solution for analyzing cellular heterogeneity and gene expression dynamics.
• The tool enhances the interpretability and robustness of single-cell clustering.
• It facilitates the integration and comparison of multi-omic and multi-modal single-cell data.