eCV {eCV}

Assess Reproducibility via Enhanced Coefficient of Variation

Description

This function estimates an "enhanced" coefficient of variation (eCV) to measure the likelihood of an omic feature being reproducible. The eCV is calculated as |SD^2 - Mean^2| / Mean^2, a metric that decreases with noise among replicates and increases with the mean intensity.

Usage

eCV(x, max.ite = 10000, n_threads = 1)

Arguments

x

A numeric matrix with rows representing omic features and columns representing sample replicates. Numeric values should be positive and reflect significance (not necessarily p-values).

max.ite

Number of samples from the null distribution (numeric). Defaults to 1e4.

n_threads

Number of threads for parallel computing. Numeric. Defaults to 1.

Details

Inferences are made based on the probabilities of eCV values originating from the group of reproducible features. It assumes that reproducible features follow a prior Normal distribution with dimension r (number of replicates). Pseudo replicates are sampled using a Probabilistic Bootstrap, assuming that the global mean vector and variance-covariance matrix across features are close to the prior's hyperparameters. eCV is computed for each random sample. The proportion of times the observed eCV is lower than or equal to the eCV from random samples is then taken as the probability of the omic feature belonging to the group of reproducible features.

Value

Returns a list with two elements:

• ecv: Numeric vector with the estimated eCV values for each omic feature.

• post_prob: Posterior probability values.

Examples

library(eCV)
set.seed(42)
# Simulate data.
out <- simulate_data(scenario = 1,n_features=1e3)

# Run eCV
ecv_out <- eCV(x = out$sim_data, max.ite = 100)


# Plot results.
library(tidyverse)

out$sim_data %>% 
as.data.frame() %>% 
mutate(`eCV Prob` = ecv_out$post_prob) %>%
ggplot(aes(x = `Rep 1`, y = `Rep 2`, color = `eCV Prob`)) +
 geom_point(size = 1) + 
 scale_color_gradientn(colors=c( "#009CA6", "#D5DADD", "#F4364C"))+ 
 theme_classic()