mrep_assessment {eCV}

Multi-replicate Reproducibility Assessment.

Description

This function wraps the different methods implemented in the package eCV to assess reproducibility of omic feature values coming from two or more sample replicates. The 'method' argument specifies any of the implemented methods: "IDR", "gIDR", "mIDR", and "eCV".

Usage

mrep_assessment(x, method = "eCV", param, n_threads = 1)

Arguments

x

A numeric matrix with rows representing the number of omic features and columns representing the number of sample replicates. The numeric values should be positive and represent significance (not necessarily p-values).

method

The name of the method used to assess reproducibility. Character. Possible values are "IDR", "gIDR", "mIDR", and "eCV". Defaults to "eCV".

param

List specifying the initial values for the parameters used by the specified method. If method is any of the IDR variants, param must be a named list with "mu", "sigma", "rho", "p", "eps", and "max.ite". If method = "eCV", param only needs "max.ite".

n_threads

Number of threads for parallel computing. Numeric. Default to 1. Only used when method is mIDR or eCV.

Details

The "IDR" method calls the traditional IDR, as implemented in the package idr (idr::est.IDR). Regardless of the number of replicates given to the function, when method="IDR", only the first two are used. Any of the other methods are meant to be used when r >= 2. Both gIDR and mIDR reduce to traditional IDR if r = 2.

Value

A list with two elements

rep_index

A numeric vector with values between zero and one, with smaller values indicating higher reproducibility

method

String storing the name of the method used

Examples

library(eCV)

# Simulate data
set.seed(42)
out <- simulate_data(scenario = 2, n_reps = 4, n_features = 1e3)

# Define parameters for each method.
params <- list(
  eCV = list(max.ite = 100),
  gIDR = list(
     mu = 2,
     sigma = 1.3,
     rho =0.8,
      p = 0.7,
    eps = 1e-3,
    max.ite = 50),
  mIDR = list(
     mu = 2,
     sigma = 1.3,
     rho =0.8,
      p = 0.7,
    eps = 1e-3,
    max.ite = 50))

# Create a list to store results
results <- NULL
methods <- c("eCV", "gIDR", "mIDR")
for (method in methods) {
  rep_index <- mrep_assessment(x = out$sim_data,
                               method = method,
                               param = params[[method]])$rep_index
  new_rows <- data.frame(value = rep_index,
                         Method = method,
                         group = out$sim_params$feature_group)
  results <- rbind(results, new_rows)
}

# Plot results

library(tidyverse)
results %>% 
mutate(group = ifelse(group == 1,"FALSE","TRUE")) %>%
ggplot(aes(x=Method, y = value,fill=group)) + 
scale_fill_manual(values = c( "#009CA6" , "#F4364C")) + 
geom_boxplot() + 
theme_classic() + 
labs(y="Reproducibility assessment", fill="Reproducible\nfeature")