Normalization method when using RnBeads to analyze MMBC (Infinium Mouse Methylation BeadChip) data

Question

[ilianna]

I am a postdoc in UCLouvain and the last year I am following courses to be able to analyze my data on my own, so I am not an expert user of R.

The last few weeks I am trying to analyze my BBMC data using RnBeads package in RStudio, but I am facing some troubles regarding the normalization process that's why I am posting here.

• I tried illumina, but I had a warning:

rnb.execute.normalization(rnb.set.filtered,   method="illumina",
                                            bgcorr.method="none")
2023-03-22 15:04:52     Inf WARNING Incompatible dataset and normalization method: not supported for Mouse Methylation Bead Chip. Changed the method to "none"

• Then I tried scaling without having warnings:

rnb.execute.normalization(rnb.set.filtered, method = "scaling",
                                              bgcorr.method = "subtraction")
2023-03-23 10:44:42     Inf  STATUS Performed background subtraction with method subtraction
2023-03-23 10:44:43     Inf  STATUS Performed normalization with method scaling

• And finally I concluded to wm.dasen, as it was the default option of normalization:

rnb.getOption("normalization.method")
[1] "wm.dasen"
rnb.getOption("normalization.background.method")
[1] "none"

rnb.execute.normalization(rnb.set.filtered, method = "wm.dasen",
                                                    bgcorr.method = "none")
2023-03-23 12:32:39     Inf  STATUS     Performed normalization with method wm.dasen

Although, I run my whole analysis with wm.dasen normalized data I am not so sure that I picked the right normalization method. Does anyone know which one is better to be selected as long as the normalization methods are not all available for mouse data?

I would appreciate your contribution.

Thank you in advance,
Ilianna

Answer 1

[mscherer]

Hi Ilianna,

Thank you for using RnBeads. For now, we have not systematically compared all the normalization methods for the Mouse array, but made very good experience with wm.dasen. I would for now recommmend to go with this normalization method, but would be also very happy to receive your feedback on the different methods.

Hope that helps,

Michael