minfi::getQC - is default badsamplecutoff of 10.5 always appropriate? ~half of samples fail by this measure

Question

[rkb965]

Hi Biostars!

I am completely new to this world -- first project question here.

I am trying to run through some QC for ewas data. I am using {minfi} and following along with the workflow (https://www.bioconductor.org/help/course-materials/2015/BioC2015/methylation450k.html#qc-plot)

About half of my samples fail by the function getQC and the default cutoff of 10.5. The samples generally look good by other metrics (eg the densityplot and densitybeanplots both look reasonably good).

1. Is the default cutoff of 10.5 almost always appropriate? Is it generally appropriate for all sample types, or is it perhaps blood-specific?

2. What might this reflect?

Thank you so much!

[Papyrus]

In that same workflow, when you do the plot of Umeth median intensity vs Meth median intensity (with plotQC(qc)), can you show the plot? It is not the same to have a single cloud of points than to have 2 clusters, one of bad quality and one of good quality.

Though from what I've seen those values are pretty low, are these frozen tissue samples? or maybe paraffin-fixed samples?

[June]

Hello,

I just started working on analyzing methylation EPIC array data using minfi. Data is generated on EPICv2 array and I'm using updated EPICv2 annotation (rgset@annotation <- c(array = "IlluminaHumanMethylationEPICv2", annotation = "20a1.hg38") Using the getQC and plotQC function. I noticed that all of samples are below 10.5 cut-off (lowest value is 8.8 and highest value is 10.48). Beta values density plot looks normal and detection p-value also looks normal. I'm confused on how should I interpret this data?

Thanks