Power Analysis for Sequencing Experiments
Core Concept
Power = probability of detecting a true effect. Underpowered studies waste resources; overpowered studies are inefficient.
RNA-seq Power Analysis
r
library(RNASeqPower) # Typical parameters # - depth: sequencing depth per sample (reads/gene) # - cv: biological coefficient of variation (0.1-0.4 typical) # - effect: fold change to detect (1.5 = 50% change) # - alpha: significance level (0.05 standard) # Calculate power for given sample size rnapower(depth = 20, n = 3, cv = 0.4, effect = 2, alpha = 0.05) # Calculate required samples for target power rnapower(depth = 20, cv = 0.4, effect = 2, alpha = 0.05, power = 0.8)
CV Guidelines
| Experiment Type | Typical CV | Notes |
|---|---|---|
| Cell lines | 0.1-0.2 | Low variability |
| Inbred mice | 0.2-0.3 | Moderate |
| Human samples | 0.3-0.5 | High variability |
| Primary cells | 0.3-0.4 | Donor-dependent |
ATAC-seq Power (ssizeRNA)
r
library(ssizeRNA)
# For differential accessibility
size.zhao(m = 10000, m1 = 500, fc = 2, fdr = 0.05, power = 0.8,
mu = 10, disp = 0.1)
Quick Reference
| Effect Size | Recommended n (CV=0.4) |
|---|---|
| 4-fold | 3 per group |
| 2-fold | 5-6 per group |
| 1.5-fold | 10-12 per group |
| 1.25-fold | 20+ per group |
Related Skills
- •experimental-design/sample-size - Detailed sample size calculations
- •experimental-design/batch-design - Accounting for batch effects in design
- •differential-expression/deseq2-basics - Running the actual DE analysis