Is there a generally accepted method used to impute missing DNA methylation data (probe beta-values)? First time dealing with this situation, would appreciate any suggestions!

Question

[Lunky]

I'm looking at DNA methylation (DNAm) data such as TCGA (e.g., BRCA, KIRP, KIRC, etc.). Currently trying to build use my model to predict DNm age on test sets, but many of the data sets are missing key probe values used in my clock/model.

I have inspected approximately 30 GEO data series and the following TCGA cancer samples: KIRP, KIRC, LUAD, LUSC, BRCA, THCA. They are each missing key probes I am using.

Is there a common way to impute the missing values in R?

[Basti]

Maybe you could have a look at the champ.impute() function from the ChAMP package : https://rdrr.io/bioc/ChAMP/man/champ.impute.html

[Marc]

I know that methylsuite (python) and Sesame (R) packages have impute techniques for missing probes. (If probes are poor quality they get removed.) But none of these imputed values are going to be super accurate without a matrix showing how all the probes are correlated across sample types. That dataset doesn't exist yet.