Benchmarking UMI-based single-cell RNA-seq preprocessing workflows.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) has advanced rapidly, with preprocessing methods crucial for generating gene count matrices.
• Numerous scRNA-seq preprocessing workflows exist, but their comparative performance and impact on downstream analyses remain understudied.

Purpose of the Study:

• To systematically benchmark 10 end-to-end scRNA-seq preprocessing workflows.
• To evaluate the impact of these workflows on gene quantification, normalization, and cell clustering.
• To guide users in selecting appropriate scRNA-seq preprocessing tools.

Main Methods:

• Benchmarking of 10 scRNA-seq preprocessing workflows: Cell Ranger, Optimus, salmon alevin, alevin-fry, kallisto bustools, dropSeqPipe, scPipe, zUMIs, celseq2, and scruff.
• Utilized datasets from CEL-Seq2 and 10x Chromium platforms with varying biological complexity.
• Assessed quantification accuracy and downstream analysis performance (normalization, clustering) using ground truth cell type labels.

Main Results:

• Significant variation observed in gene detection and quantification across different scRNA-seq preprocessing workflows and datasets.
• Despite workflow differences, downstream analysis using robust normalization and clustering methods yielded highly concordant cell type clustering results.
• The choice of preprocessing workflow had a less pronounced impact on final clustering outcomes compared to subsequent analysis steps.

Conclusions:

• The selection of a specific scRNA-seq preprocessing workflow is less critical than other analytical steps for achieving accurate cell type classification.
• This comprehensive comparison provides valuable insights into the characteristics of common scRNA-seq preprocessing tools.
• Findings aid researchers in navigating the landscape of scRNA-seq analysis pipelines and making informed decisions.