Power analysis for microRNA sequencing experiments helps determine the sample size required to detect differential expression with sufficient statistical power. This process is important because microRNA data often exhibit high variability, low expression levels, and many zero counts, which complicate analytical calculations.

Simulation-based methods are recommended over simple formulas, as they better account for the characteristics of sequencing data. One effective tool is the powsimR package in R, which supports power estimation for bulk RNA sequencing and can be adapted for microRNA sequencing.

To use powsimR, first install it:

install.packages("devtools")
devtools::install_github("bvieth/powsimR", build_vignettes = TRUE)

Then, perform a simulation using your microRNA count data:

library(powsimR)

# Estimate parameters from your count matrix
params <- estimateParam(countData = your_count_matrix,
                        distribution = "NB",
                        RNAseq = "bulk",
                        normalisation = "TMM")

# Simulate differential expression for various sample sizes
simRes <- simulateDE(params,
                     n = c(3, 6, 9, 12),
                     DEmethod = "DESeq2",
                     nGroups = 2)

# Evaluate power at FDR 0.05
pow <- evaluateDE(simRes, alpha.type = "fdr", alpha.nominal = 0.05)
plotPower(pow)

This approach simulates negative binomial distributions typical of microRNA counts and evaluates power using DESeq2.

For a simpler analytical method, use the RNASeqPower package:

library(RNASeqPower)

# Calculate samples needed (per group) for given parameters
rnapower(depth = 10, cv = 0.4, effect = 1.5, alpha = 0.05, power = 0.8)

Here, depth is average reads per microRNA, cv is coefficient of variation, and effect is fold change.

Studies indicate that microRNA sequencing often requires larger sample sizes than messenger RNA sequencing due to variability; for example, one analysis suggested at least 17 replicates per group for reliable detection.

Refer to these resources for further details:

• Power and sample size calculations for high-throughput sequencing data (https://academic.oup.com/bib/article/19/6/1247/3865509)
• Harmonization of quality metrics and power calculation in multi-omic studies (https://www.nature.com/articles/s41467-020-16937-8)

Kevin