Quantification and statistical modeling of droplet-based single-nucleus RNA-sequencing data

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Single-nucleus RNA sequencing (snRNA-seq) is crucial for analyzing gene expression in complex tissues where cell dissociation is challenging.
• Previous studies established that droplet-based single-cell RNA sequencing (scRNA-seq) data are not zero-inflated.
• The distributional properties of droplet-based snRNA-seq data have not been systematically evaluated.

Purpose of the Study:

• To determine the probability distributions of droplet-based snRNA-seq data.
• To assess the transferability of statistical models from scRNA-seq to snRNA-seq.
• To investigate the impact of quantification strategies and reference transcriptome choices on snRNA-seq data analysis.

Main Methods:

• Analysis of pseudonegative control data from mouse cortex and kidney samples sequenced using 10x Genomics Chromium and DropSeq systems.
• Evaluation of droplet-based snRNA-seq data using negative binomial distribution models.
• Comparison of snRNA-seq data processed with reference transcriptomes including and excluding intronic regions.

Main Results:

• Droplet-based snRNA-seq data adhere to a negative binomial distribution, supporting the application of parametric models used for scRNA-seq.
• Quantification strategies significantly influence downstream analyses and cell type classification.
• Reference transcriptomes lacking intronic regions lead to reduced library sizes and inaccurate cell type identification.
• A gene length bias was confirmed in snRNA-seq data, affecting both exonic and intronic reads.

Conclusions:

• Parametric statistical models validated for scRNA-seq are applicable to snRNA-seq data.
• The choice of reference transcriptome, particularly the inclusion of intronic regions, is critical for accurate snRNA-seq analysis and biological interpretation.
• Further investigation into the causes of gene length bias in snRNA-seq is warranted.