Minnow: a principled framework for rapid simulation of dscRNA-seq data at the read level

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• High-throughput single-cell RNA sequencing (scRNA-seq) generates vast gene expression data, driving demand for sophisticated analysis tools.
• Current scRNA-seq analysis pipelines often rely on simulated count matrices or known marker genes for validation, overlooking sequence-level processing effects.
• The impact of read alignment and unique molecular identifier (UMI) deduplication methods on scRNA-seq data quality is not well understood.

Purpose of the Study:

• To introduce minnow, a novel simulation framework for droplet-based single-cell RNA sequencing (dscRNA-seq) experiments.
• To model sequence-level characteristics, including PCR amplification, cellular barcodes (CB), UMI selection, fragmentation, and sequencing.
• To assess the performance of common scRNA-seq quantification pipelines using realistic, sequence-level simulated data.

Main Methods:

• Developed minnow, a comprehensive simulation framework for dscRNA-seq experiments.
• Incorporated sequence-level features such as PCR amplification, CB and UMI processing, and fragmentation.
• Modeled gene-level ambiguity characteristics observed in real scRNA-seq data.

Main Results:

• Evaluated the performance of standard gene-by-cell count matrix generation pipelines using simulated dscRNA-seq data.
• Demonstrated the significant impact of sequence-level ambiguity on accurate gene quantification.
• Illustrated minnow's utility in evaluating different quantification pipeline outputs.

Conclusions:

• Minnow provides a realistic simulation of dscRNA-seq experiments, capturing crucial sequence-level details.
• Sequence-level variations significantly influence the accuracy of gene quantification in scRNA-seq data.
• This framework aids in assessing and improving scRNA-seq data processing and analysis pipelines.