Missing data and technical variability in single-cell RNA-sequencing experiments.

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• High-throughput gene expression technologies like RNA-Sequencing (RNA-seq) traditionally required large cell numbers.
• Recent advances enable genome-wide gene expression measurement at the single-cell level using single-cell RNA-Sequencing (scRNA-seq).
• scRNA-seq data exhibits unique characteristics, including a high proportion of zero expression values and substantial cell-to-cell variability.

Purpose of the Study:

• To investigate the extent to which technical variation drives the observed cell-to-cell variability in scRNA-seq data.
• To identify and quantify systematic errors, such as batch effects, in scRNA-seq datasets.
• To demonstrate how technical artifacts can be misinterpreted as biological findings.

Main Methods:

• Analysis of published scRNA-seq datasets through an assessment experiment.
• Examination of the proportion of zero expression values in scRNA-seq data.
• Evaluation of the impact of technical variation and batch effects on gene expression variability.

Main Results:

• scRNA-seq produces a higher proportion of zero expression values than expected, indicating technical bias.
• This technical bias disproportionately affects lower expressed genes.
• Observed cell-to-cell variability in scRNA-seq data is substantially explained by systematic technical errors.
• Batch effects and confounded experiments significantly exacerbate these technical issues.

Conclusions:

• A significant portion of zero expression counts in scRNA-seq data is attributable to technical variation, not biological absence of expression.
• Technical variability, including batch effects, can be mistaken for novel biological insights if not properly addressed.
• Careful experimental design and data normalization are crucial for accurate interpretation of scRNA-seq results.