Expression-based species deconvolution and realignment removes misalignment error in multispecies single-cell data.

Area of Science:

• Genomics
• Bioinformatics
• Single-cell analysis

Background:

• Single-cell RNA sequencing (scRNA-seq) is widely used for xenograft samples.
• A comprehensive bioinformatics pipeline for human-mouse mixed scRNA-seq is lacking.
• Homologous genes between human and mouse necessitate evaluation of misalignment errors.

Purpose of the Study:

• To develop a bioinformatics pipeline for accurate analysis of mixed human and mouse single-cell data.
• To assess and minimize misalignment and multi-mapping errors in combined species reference data.
• To address false-positive signals from ambient RNA in multispecies scRNA-seq.

Main Methods:

• Developed a novel bioinformatics pipeline with expression-based species deconvolution.
• Assessed misalignment errors and exonic multi-mapping events using combined human and mouse reference data.
• Evaluated ambient RNA contamination from the other species.

Main Results:

• Misalignment errors using combined reference data averaged 0.78% of total reads, affecting specific genes.
• The developed pipeline demonstrated superior clustering and batch effect mitigation compared to Cell Ranger.
• Successfully applied the pipeline to multispecies xenograft tissue, identifying samples and capturing diverse tumor microenvironment cell types.

Conclusions:

• A robust bioinformatics pipeline for mixed human and mouse single-cell data is presented.
• The pipeline is applicable to pooled libraries for cost-effective single-cell data generation.
• Misalignment, multi-mapping errors, and ambient RNA are critical considerations for multispecies scRNA-seq analysis.