PROPER

PROPER performs prospective power assessment for differential expression analysis of RNA-seq data by using semi-parametric simulation to model baseline expression, biological variation, and differential expression patterns while evaluating stratified power and false discovery cost.

Key Features:

• Prospective power assessment: Assesses statistical power prospectively for RNA-seq differential expression (DE) studies rather than performing direct sample size calculations.
• Semi-parametric simulation: Generates synthetic RNA-seq datasets that mirror experimental conditions and flexibly model baseline expressions, biological variation, and DE patterns.
• Fraction of DE genes: Accounts explicitly for the fraction of genes that are differentially expressed when evaluating power.
• Expression magnitude distribution: Models the distribution of expression fold changes to reflect DE magnitude variability.
• Overall gene expression levels: Incorporates baseline overall gene expression levels into simulated data and power evaluations.
• Sequencing coverage effects: Evaluates the impact of sequencing coverage (depth) on detection power.
• Type I error control strategies: Assesses the effects of different type I error control strategies on power and false discoveries.
• Stratified power and false discovery cost: Computes stratified power metrics and false discovery cost to provide nuanced evaluation of detection performance.
• Design guidance: Provides quantitative guidance for sample size determination, sequencing depth optimization, and gene filtering strategies in DE study planning.
• Implementation: Implemented as an open-source R package for computational assessment of RNA-seq DE power.

Scientific Applications:

• RNA-seq experimental design: Planning and comparing alternative RNA-seq study designs based on expected detection power.
• Sample size determination: Estimating required sample sizes to achieve target power for DE detection under specified assumptions.
• Sequencing depth optimization: Quantifying how different sequencing coverage levels affect sensitivity to detect DE genes.
• Gene filtering strategy evaluation: Assessing the impact of gene filtering choices on power and false discovery outcomes.
• Multiple testing and error control assessment: Evaluating how type I error control strategies influence power and false discovery trade-offs.

Methodology:

Semi-parametric simulation of synthetic RNA-seq datasets that model baseline expression, biological variation, and DE patterns, followed by power assessment computing stratified power and false discovery cost.