Benchmarking single cell RNA-sequencing analysis pipelines using mixture control experiments.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) is rapidly advancing, generating numerous data analysis methods.
• A lack of standardized benchmark datasets hinders systematic comparison of these scRNA-seq analysis tools.

Purpose of the Study:

• To create a realistic benchmark experiment for evaluating scRNA-seq data analysis methods.
• To provide a comprehensive framework for comparing common scRNA-seq analysis steps.

Main Methods:

• Generated 14 scRNA-seq datasets using droplet and plate-based protocols with single cells and 'pseudo cells' from five cancer cell lines.
• Compared 3,913 combinations of data analysis methods for normalization, imputation, clustering, trajectory analysis, and data integration.

Main Results:

• Identified optimal data analysis pipelines tailored to specific scRNA-seq data types and analysis tasks.
• Demonstrated that method performance varies significantly depending on the task and data characteristics.

Conclusions:

• The developed benchmark datasets and analysis framework facilitate rigorous evaluation of scRNA-seq methods.
• This resource aids researchers in selecting appropriate computational strategies for their scRNA-seq experiments.