A comparison of integration methods for single-cell RNA sequencing data and ATAC sequencing data.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell genomics offers insights into cellular phenotypes and genetics.
• Emerging technologies enable multimodal single-cell data collection (e.g., transcriptomes, epigenomes).
• Integrating these diverse datasets for cell correspondence remains a significant challenge.

Purpose of the Study:

• To systematically review popular single-cell integration methods.
• To evaluate over 10 integration methods on gold-standard datasets.
• To discuss limitations, applications, and future directions for multimodal single-cell data integration.

Main Methods:

• Systematic literature review of single-cell integration methods.
• Evaluation of methods for cell label transfer, data visualization, and clustering.
• Benchmarking popular integration methods on paired and unpaired datasets.

Main Results:

• Identified popular single-cell integration methods and models.
• Evaluated method performance across various downstream tasks.
• Assessed data preferences, limitations, and applications of different integration approaches.

Conclusions:

• Data integration at a biologically relevant granularity is crucial.
• Balancing biological discovery with noise reduction requires accounting for modality discrepancies.
• The review provides a comprehensive guide to selecting and applying single-cell integration methods.