Benchmarking of computational demultiplexing methods for single-nucleus RNA sequencing data

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Single-nucleus RNA sequencing (snRNA-Seq) offers deep insights into cellular heterogeneity and gene expression in complex tissues.
• High costs associated with snRNA-Seq limit the number of samples that can be analyzed.
• Sample pooling and subsequent demultiplexing using genetic variants present a cost-effective solution for increasing sample throughput.

Purpose of the Study:

• To comprehensively benchmark leading software tools for genetic variant-based sample demultiplexing in snRNA-Seq.
• To compare variant calling from SNP array and bulk RNA-Seq data for demultiplexing.
• To evaluate the impact of doublet percentages and variant calling tools on demultiplexing performance.

Main Methods:

• Benchmarking of Vireo, Souporcell, Freemuxlet, and scSplit software for sample demultiplexing using genetic variants.
• Comparison of genetic variant data derived from SNP array (gDNA) and sample-matched bulk RNA-Seq.
• Evaluation using simulated multiplexed datasets (2, 4, 6 samples; 0-30% doublets) and validation with sex-linked genes.

Main Results:

• All tested tools, except scSplit, achieved high recall and precision (80-85% accuracy), with Vireo demonstrating the best performance.
• Demultiplexing accuracy was influenced by the choice of variant calling tool and decreased with increasing doublet percentages.
• Successful deployment of demultiplexing on real-world 10x RNA-Seq data from human heart and cross-species samples was demonstrated.

Conclusions:

• Vireo is recommended as the top-performing software for genetic variant-based sample demultiplexing in snRNA-Seq.
• The choice of variant calling method impacts demultiplexing tool performance, highlighting the need for careful consideration.
• Demultiplexing strategies enhance cost-efficiency, improve doublet detection, and facilitate analysis of pooled snRNA-Seq data.