Quantify and Control Reproducibility in High-Throughput Experiments

Estimate the proportions of the null and the reproducibility and non-reproducibility of the signal group for the input data set. The Bayes factor calculation and EM (Expectation Maximization) algorithm procedures are also included.

Installation

library(devtools)
install_github("ArtemisZhao/INTRIGUE/R_src")

Example dataset

library(INTRIGUE)
data("heterodata")

This is a simulated datasets with n=5000 experimental units (e.g., genes) and m=2 replicates. The true proportions for the null, the reproducible and the irreproducible signal categories are 0.80, 0.18, 0.02, respectively.

CEFN prior

##for convenience, only test on first 100 units 
hetero.out.cefn<-hetero(heterodata[1:100,],fdr.level=0.05)

META prior

hetero.out.meta<-hetero(heterodata[1:100,],use_cefn=FALSE,fdr.level=0.05)

Outcome

names(hetero.out.cefn)

## overall proportion
print(hetero.out$est_prop)

Return a list containing the following,

1. gridweight: estimated weight on each grid point.

2. ind_prob: individual probability of falling into three groups.

3. est_prop: estimated overall proportion for three groups.

4. lfdr: false discovery rate.

5. significant: decision based on the fdr.level in the input.