Hi
Our lab sent a number of samples off to a contractor for Illumina EPIC Methylation assay, and one of the reports returned failures of 40% of the samples due to their being below a certain threshold of unmethylated to methylated intensities (log2 of the median across all CpG loci per sample taken).
I'd like to ask about this threshold, because 40% of samples is very many to lose for the analysis. I have been searching for links on this. While there are other QC criteria, this Log2 median unmeth/meth intensity is very punishing. Is there anybody with guidance on this measure?
I was not able to reproduce the exact values. I used raw values for the green and red channels as intensity, and though it was close to the contractors values, it was still different, and the contractor's threshold is not the one I should use (many more of my sample passed in my reproduction).
Any suggestions or simple links welcome.
Many thanks!
Hi, I understand this is very disturbing. Do you know which threshold did the contractor applied ? It could be a nice if you could perform the first few steps of QC as recommended in the minfi tutorial : https://www.bioconductor.org/help/course-materials/2015/BioC2015/methylation450k.html#qc-plot After plotting plotQC(qc), you should obtain something like this and we should be able to see what is going on there:
Thanks !
Thanks for the comment. the threshold is y=21-x, it can be worked out from the graph you posted. The contractor used the same one. In the example you post, there are no failing samples, so the issue does not arise, and consequently it's not considered. I cannot find why that threshold is used.