scPower

scPower provides power analysis and experimental design optimization for cell type-specific interindividual differential gene expression (DE) and expression quantitative trait loci (eQTL) studies using single-cell RNA sequencing (RNA-seq).

Key Features:

• Power Calculation: Computes statistical power to detect cell type-specific inter-individual DE and eQTL signals from single-cell RNA-seq data.
• Optimization for Budget Constraints: Optimizes combinations of sample size, number of cells per individual, and sequencing depth under budget constraints to maximize detection power.
• Customizable Experimental Priors: Allows input of experimental priors (effect sizes and expression distributions) from example datasets or by estimation from new datasets.
• Modeling Relationships: Models the relationship between sample size, number of cells per individual, and sequencing depth and their impact on power to detect DE genes within cell types.
• Platform Evaluation: Evaluated across multiple single-cell profiling platforms and tissues using both unique molecular identifier (UMI) counts and read counts.
• Implementation: Implemented as an R package.

Scientific Applications:

• Cell type-specific DE studies: Designs and assesses power for inter-individual differential gene expression analyses at cell type resolution using single-cell RNA-seq.
• eQTL mapping: Informs sample size and sequencing strategies for expression quantitative trait loci (eQTL) studies in single-cell transcriptomics.
• Experimental design trade-offs: Guides decisions on sequencing depth versus number of cells per individual for budget-constrained single-cell transcriptomic experiments.

Methodology:

Implements a statistical framework that systematically evaluates optimal parameter combinations for multi-sample single-cell transcriptomic experiments and models the impact of sample size, cells per individual, and sequencing depth on power; evaluations across single-cell technologies and tissues using UMI and read counts indicate that shallow sequencing of many cells often yields higher power than deep sequencing of fewer cells.