18 months ago by
University College London
Even EPIC array have different versions. For example, on Illumina website, you could find various versions of EPIC manifests. Version 2 have different number of CpG covered with version 4 if I remember correctly.
Different versions of EPIC data would have slightly differences on CpG covered, but should now be much. Thus, a proper way to address this problem is extracting common CpGs in these samples.
However, I still suggest you check your samples IDAT data, you just need to us bath command to list their file size, then you should be able to find differences between samples. Then, after loading, I suggest you load this information into pd as a covariate, then see if it has system bias to final data set.