Error using preprocessNoob (minfi) for processing of Illumina array data

Question

[steffanchristiansen]

I am trying to run a script for data processing (https://github.com/allanjust/methylation-lab/blob/master/code/meth01_process_data.R) of Infinium MethylationEPIC BeadChip (Illumina) data. However, I am unable to run the preprocessing step of the RGChannelSet (WB) using the preprocessNoob function (minfi package) and gets following error:

preprocessNoob(WB) Error in huber(controls[, i]) : cannot estimate scale: MAD is zero for this sample

I have removed one of the samples from the RGChannelSet which failed to pass all the control parameters (NaN or Infinity) from the BeadArray Controls Reporter software. However, it did not affect the error.

Any suggestions for solving the issue?

Thank you, Steffan

[jamesebarrett]

Hi there, I'm getting the same error with my EPIC array data. Did you manage to resolve it in the end?

[steffanchristiansen]

Yes, I did. Apparently, two of the samples caused the error. The other sample had no "NaN" or "Infinity values" but the values of BisulfiteConversion1Red, NonPolymorphicGreen and NonPolymorphicRed was way lower than the remaining samples (1.0, 2.2 and 1.2, respectively).

Removing both samples from the RGChannelSet solved the issue.