mIDR {eCV}

Estimate meta Irreproducible Discovery Rate (mIDR).

Description

This function extends the Li et al. (2011) copula mixture model, originally implemented in idr::est.IDR, to accommodate any number of replicates. It computes the local IDR for all pairwise combinations of replicates. Then it computes a "meta" local IDR score using the formula: 1 - (1 - idr_1)*...*(1 - idr_C(r,2)), where C(r,2) represents the number of all pairwise combinations of scores. Once the meta local IDR is obtained, the expected IDR scores are obtained in the same way as in the traditional IDR procedure.

Usage

mIDR(x, mu, sigma, rho, p, eps = 0.001, max.ite = 20, 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).

mu

Starting value for the mean of the reproducible component Numeric.

sigma

Starting value for the standard deviation of the reproducible component Numeric.

rho

Starting value for the correlation coefficient of the reproducible component Numeric.

p

Starting value for the proportion of the reproducible component Numeric.

eps

Stopping criterion. Iterations stop when the increment of the log-likelihood is less than "eps" times the log-likelihood. Defaults to 0.001.

max.ite

Maximum number of iterations. The default is 30.

n_threads

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

Value

Returns a list of two elements:

idr

A numeric vector of the local meta IDR for each observation.

IDR

A numerical vector of the expected meta IDR for observations that are as irreproducible or more irreproducible than the given observations.

References

Q. Li, J. B. Brown, H. Huang, and P. J. Bickel. (2011) Measuring reproducibility of high-throughput experiments. Annals of Applied Statistics, Vol. 5, No. 3, 1752-1779.

Examples

library(eCV)
set.seed(42)

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

# Set initial parameter values.
mu <- 2
sigma <- 1.3
rho <- 0.8
p <- 0.7

# Get meta local IDR scores.
midr_out <- mIDR(x = out$sim_data, mu, sigma, rho, p)

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