Interpretation of control strip and beta density plots for Illumina methylation array data using minFi

Question

[sichan]

Hello,

Using the minFi R package, I created the following control strip plot and beta density plot:

For the control strip plot on the left, from what I've read, by default, each of the EPIC control probe types (e.g bisulfite conversion I, conversion, extension, etc) will have its own plot. Each row represents a sample and each red and green point represents the unmethylated and methylated signal intensity, respectively, for a probe of that control type. My understanding is that a good result is when the distribution of values for each of those control probes are similar between red and green points. This indicates that there are no biases that result in the unmethylated or methylated signal being more readily detected than the other. Is this the correct interpretation? If that is the case, then I seem to have some bias because the methylated data points seem to have a greater intensity than the unmethylated points.

Regarding the density plot on the right, I noticed that there seem to be two peaks at the lower end. A beta-value at ~0 and another one at ~0.1. Does anyone know what this extra peak indicates?

Thank you.

[steffanchristiansen]

Did you find an explanation for your observed pattern? I observe a similar pattern in a dataset where half of the samples have the "double peaks" while the other samples havn't. The double peaks are located at ~0 and ~0.2.

Answer 1

[ISB]

My understanding is that the densityplot gives us a per-sample distribution of avarage beta values- The expected distribution is bimodal with the two peaks that you have in you figure, representing unmethylated and methylated signals. I expect the density plot is of raw data? I would be a little concerend about your samples that peak in the centre. Hopefully the density plot is cleaner after QC steps.