Normalised methylation data - genetics-of-dna-methylation-consortium/godmc_phase2 GitHub Wiki

DNA methylation data pre-processing

This page outlines the steps that need to be performed prior to running the GoDMC pipeline.

Option 1 (preferred)

Ideally, we would prefer it if you normalise and QC your data using the R meffil package. Please make sure you have installed version 1.3.8 or higher. See here for installation instructions. You can check the version like this:

packageVersion("meffil")

Meffil has been optimised for speed and memory, and instructions on how to do this can be found here:

  1. Perform sample QC
  2. Normalise QC'd samples

Option 2

While we our prefer you follow our recommended normalisation approach, we do not enforce this. If not meffil, then we would prefer that you would use an alternative functional normalisation method. If you prefer to use an alternative normalisation method then please ensure that the DNA methylation data you provide meets the following criteria:

  • Beta values (i.e. proportions) should be used to quantify DNA methylation level.
  • The methylation data is an R numeric matrix object where each row is a CpG site and each column is a sample.
  • The rownames must be the unique cg identifiers and the column names are the IDs that correspond to the samples, and that match to sample IDs in the other datasets (genetic, covariate etc).

Data format

Regardless which normalisation method you use the following need to be adhered to:

  • You should avoid spaces in your file name.
  • You should keep sites from chrX and chrY.
  • You should exclude the SNP probes with IDs that start with rsXXXX (65 on the 450K array and 59 on the EPIC array).
  • You should keep nonspecific binding probes, probes with SNPs in their sequence, multimapping probes. We will filter out these probes at the end.
  • Underscores should not be used in the sample IDs.
  • The pipeline removes outliers with the Tukey method in the processing methylation module so you shouldn't have NAs in your methylation dataset.
  • The methylation matrix should be saved as a .RData file (each row is a CpG site and each column is a sample) and the methylation matrix object should be called norm.beta. e.g. in R:
save(norm.beta, file="/path/to/godmc/input_data/methylation.RData")