A multicenter study benchmarking single-cell RNA sequencing technologies using reference samples.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Comparing diverse single-cell RNA sequencing (scRNA-seq) datasets from different technologies and labs is challenging.
• Accurate biological interpretation requires guidance on selecting appropriate bioinformatic methods for varied data types and platforms.

Purpose of the Study:

• To compare scRNA-seq platforms and bioinformatic methods for data integration.
• To provide practical guidance for optimizing platform and software selection in scRNA-seq studies.

Main Methods:

• Comparison of different scRNA-seq platforms using reference samples (breast cancer cells, B cells).
• Evaluation of various preprocessing, normalization, and batch-effect correction methods across multiple centers.
• Assessment of dataset characteristics (heterogeneity, platform) impact on bioinformatic method performance.

Main Results:

• Batch-effect correction significantly impacts cell classification accuracy, more than preprocessing or normalization.
• Dataset characteristics critically influence the choice of optimal bioinformatic methods.
• High reproducibility across centers and platforms is achievable with appropriate bioinformatic strategies.

Conclusions:

• Optimal bioinformatic method selection is dependent on scRNA-seq dataset characteristics.
• Batch-effect correction is the most critical step for accurate cell classification in integrated scRNA-seq data.
• This study provides essential guidance for researchers designing scRNA-seq experiments to ensure data integrity and reproducibility.