Transcriptome-wide analysis of differential expression in perturbation atlases

Area of Science:

• Genomics
• Computational Biology
• Systems Biology

Background:

• Single-cell CRISPR screens like Perturb-seq offer large-scale transcriptomic profiling of genetic perturbations.
• However, the resulting data are often noisy, leading to the undetected effects of many perturbations.

Purpose of the Study:

• Introduce transcriptome-wide analysis of differential expression (TRADE), a novel statistical model.
• To accurately account for estimation error and identify true differential expression effects.
• To quantify the total effect of a perturbation across the transcriptome, termed 'transcriptome-wide impact'.

Main Methods:

• Developed TRADE, a statistical model designed for the distribution of differential expression effects.
• Applied TRADE to analyze several large Perturb-seq datasets.
• Estimated the transcriptome-wide impact of gene perturbations.

Main Results:

• TRADE detects numerous transcriptional effects missed by standard analyses, revealing them in aggregate.
• A typical gene perturbation impacts an estimated 45 genes; an essential gene impacts over 500.
• Observed moderate consistency of perturbation effects across different cell types.
• Identified perturbations with dose-dependent qualitative variations in transcriptional responses.
• Clarified the relationship between genetic and transcriptomic correlations in neuropsychiatric disorders.

Conclusions:

• TRADE enhances the detection of subtle but widespread transcriptional effects from noisy single-cell CRISPR screen data.
• The transcriptome-wide impact metric provides a comprehensive measure of perturbation effects.
• Findings advance the understanding of gene function, regulatory networks, and disease mechanisms.