JOINTLY: interpretable joint clustering of single-cell transcriptomes.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell and single-nucleus RNA sequencing (sxRNA-seq) is vital for understanding cellular states.
• Technical variation in sxRNA-seq can obscure true biological signals.
• Integrating data across different batches is crucial but challenging.

Purpose of the Study:

• To develop a novel algorithm, JOINTLY, for joint clustering of sxRNA-seq datasets across batches.
• To improve the accuracy and interpretability of sxRNA-seq data integration.
• To construct a comprehensive reference atlas of white adipose tissue (WATLAS).

Main Methods:

• Development of the JOINTLY algorithm for batch-corrected sxRNA-seq data integration.
• Benchmarking JOINTLY against existing state-of-the-art batch integration methods.
• Application of JOINTLY to construct the WATLAS resource.

Main Results:

• JOINTLY demonstrates competitive or superior performance in clustering tasks compared to existing methods.
• The algorithm effectively integrates sxRNA-seq data while preserving subtle biological differences.
• JOINTLY facilitates cell type annotation and the identification of signaling pathways.
• The WATLAS resource characterizes four adipocyte subpopulations and maps changes in obesity.

Conclusions:

• JOINTLY is a robust and interpretable tool for sxRNA-seq data integration.
• The algorithm enhances the discovery of biological insights from multi-batch scRNA-seq studies.
• WATLAS provides a valuable community resource for white adipose tissue research, particularly concerning metabolic disease.