Irreproducible Discovery Rate (IDR)

The IDR (Irreproducible Discovery Rate) framework is a unified approach to measure the reproducibility of findings identified from replicate experiments and provide highly stable thresholds based on reproducibility. Unlike the usual scalar measures of reproducibility, the IDR approach creates a curve, which quantitatively assesses when the findings are no longer consistent across replicates. In layman's terms, the IDR method compares a pair of ranked lists of identifications (such as ChIP-seq peaks). These ranked lists should not be pre-thresholded i.e. they should provide identifications across the entire spectrum of high confidence/enrichment (signal) and low confidence/enrichment (noise). The IDR method then fits the bivariate rank distributions over the replicates in order to separate signal from noise based on a defined confidence of rank consistency and reproducibility of identifications i.e the IDR threshold.

The method was developed by Qunhua Li and Peter Bickel's group and is extensively used by the ENCODE and modENCODE projects and is part of their ChIP-seq guidelines and standards.

Installation

• Get the current repo

wget https://github.com/nboley/idr/archive/2.0.2.zip

• Install the dependencies

• python3
• python3 headers
• numpy
• setuptools
• matplotlib (only required for plotting the results)

In Ubuntu 14.04+ one can run: (sudo) apt-get install python3-dev python3-numpy python3-setuptools python3-matplotlib

In a shared environment, the dependencies and idr package may need to be installed locally. Anaconda largely automates this process. To install anaconda, which includes all the neessary dependencies:

Download [Anaconda3-2.2.0-Linux-x86_64.sh](http://continuum.io/downloads#py34) 
bash Anaconda3-2.2.0-Linux-x86_64.sh

• Download and unzip the idr code

wget https://github.com/nboley/idr/archive/2.0.2.zip
unzip 2.0.2.zip
cd 2.0.2/

• Then install idr

python3 setup.py install

Usage

List all the options

idr -h

Sample idr run using test peak files in the repo

idr --samples ../idr/test/data/peak1 ../idr/test/data/peak2

Run idr using an oracle peak list (e.g. peaks called from merged replicates):

idr --samples ../idr/test/data/peak1 ../idr/test/data/peak2 --peak-list ../idr/test/data/merged_peaks

Peak matching

The method in which peaks are matched can significantly affect the output. We have chosen defaults that we believe are reaosnable in the vast majoroity of cases, but it may be worth exploring the various options for your data set.

• --peak-list is not provided

Peaks are grouped by overlap and then merged. The merged peak aggregate value is determined by --peak-merge-method.

Peaks that don't overlap another peak in every other replicate are not included unless --use-nonoverlapping-peaks is set.

• --peak-list is provided

Peaks are grouped by overlap, and then for each oracle peak a single peak from each replicate is chosen that overlaps the oracle peak. If there are multiple peaks that overlap the oracle, then ties are broken by applying the following criteria in order: 1) choose the replicate peak with a summit closest to the oracle peak's summit 2) choose the replicate peak that has the largest overlap with the oracle peak 3) choose the replicate peak with the highest score

Output

Output file format

The output format mimics the input file type, with some additional fields.

We provide an example for narrow peak files - note that the first 6 columns are a standard bed6, the first 10 columns are a standard narrowPeak. Also, for columns 7-10, only the score that the IDR code used for rankign will be set - the remaining two columns will be set to -1.

Broad peak output files are the same except that they do not include the the summit columns (e.g. columns 10, 18, and 22 for samples with 2 replicates)

1. chrom string
   Name of the chromosome for common peaks

2. chromStart int
   The starting position of the feature in the chromosome or scaffold for common peaks, shifted based on offset. The first base in a chromosome is numbered 0.

3. chromEnd int
   The ending position of the feature in the chromosome or scaffold for common peaks. The chromEnd base is not included in the display of the feature.

4. name string
   Name given to a region (preferably unique) for common peaks. Use '.' if no name is assigned.

5. score int
   Contains the scaled IDR value, min(int(log2(-125IDR), 1000). e.g. peaks with an IDR of 0 have a score of 1000, idr 0.05 have a score of int(-125log2(0.05)) = 540, and idr 1.0 has a score of 0.

6. strand [+-.] Use '.' if no strand is assigned.

7. signalValue float
   Measurement of enrichment for the region for merged peaks

8. p-value float
   Merged peak p-value

9. q-value float
   Merged peak q-value

10. summit int
    Merged peak summit

11. localIDR float -log10(Local IDR value)

12. globalIDR float -log10(Global IDR value)

13. rep1_chromStart int
    The starting position of the feature in the chromosome or scaffold for common replicate 1 peaks, shifted based on offset. The first base in a chromosome is numbered 0.

14. rep1_chromEnd int
    The ending position of the feature in the chromosome or scaffold for common replicate 1 peaks. The chromEnd base is not included in the display of the feature.

15. rep1_signalValue float
    Signal measure from replicate 1. Note that this is determined by the --rank option. e.g. if --rank is set to signal.value, this corresponds to the 7th column of the narrowPeak, whereas if it is set to p.value it corresponds to the 8th column.

16. rep1_summit int
    The summit of this peak in replicate 1.

[rep 2 data]

...

[rep N data]

Plot output

Upper Left: Replicate 1 peak ranks versus replicate 2 peak ranks - peaks that do not pass the specified idr threshold are colered red.

Upper Right: Replicate 1 log10 peak scores versus replicate 2 log10 peak scores - peaks that do not pass the specified idr threshold are colered red.

Bottom Row: Peaks rank versus idr scores are plotted in black. The overlayed boxplots display the distribution of idr values in each 5% quantile. The idr values are thresholded at the optimization precision - 1e-6 bny default.

Contributors

The main contributors of IDR code:

• Nathan Boleu - Kundaje Lab, Dept. of Genetics, Stanford University
• Anshul Kundaje - Assistant Professor, Dept. of Genetics, Stanford University
• Peter J. Bickel - Professor, Dept. of Statistics, University of California at Berkeley

References

"Measuring reproducibility of high-throughput experiments" (2011), Annals of Applied Statistics, Vol. 5, No. 3, 1752-1779, by Li, Brown, Huang, and Bickel

Issues

If you notice any problem with the code, please file an issue over here