Distributed ChAMP pipeline - distChAMP

This is a simple tool to create myNorm file from a large amount of data [*.idats files] using ChAMP package [1]. Instead of processing all files simultaneously the tool uses distributed manner. This strategy requires less RAM memory at the cost of computation time, and has no significant influence on generated beta-values. For example 30GB of RAM memory is not sufficient to process 650 samples [EPIC + BMIQ] using standard ChAMP pipeline, however using distChAMP it is not a challenge.

Please note that final myNorm object still must fit into RAM memory.

How does it work?

Sample sheet containing n-samples is shuffled and then split into k-chunks containing ~ n/k elements. Then each chunk is processed into temporary myNorm file using ChAMP pipeline. Next all temporary myNorms are merged into final object.

How to run

To start simply clone repository:

git clone https://github.com/ClinicalEpigeneticsLaboratory/distChAMP.git

To run run_distributed_champ function you need to install all necessary packages:

Rscipt install.R

and use:

source("path_to_file.R")

when function is correctly loaded type:

run_distributed_champ(path_ss = path_to_sample_sheet_file, 
                      output = path_to_output_directory,
                      path_idats = path_to_idats_directory,
                      batch_size = number_of_samples_per_chunk)

• path_ss -> path to sample sheet file [can not be in the same directory as *.idats].
• output -> path to output directory.
• path_idats -> path to directory containing *idats.
• batch_size -> number of samples per chunk.

ChAMP args:

• array_type -> EPIC / 450K.
• norm_type -> BMIQ / SWAN / PBC / FunctionalNormalization
• cores -> number of cores to use [accelerate only normalization step]
• method -> champ or minfi

Finally output directory contains myNorm file, and QC/Norm directories generated per each chunk.

1. Tian Y, Morris TJ, Webster AP, Yang Z, Beck S, Andrew F, Teschendorff AE (2017). “ChAMP: updated methylation analysis pipeline for Illumina BeadChips.” Bioinformatics, btx513. doi: 10.1093/bioinformatics/btx513.