EPIC V2 to EPIC V1

[PierreSolomon]

Hello,

I have methylation data from the EPIC V2 array and I wish to deconvolve it (to the different PBMC cell types). My problem is that all the packages I found that deal with deconvolution of methylation data only support 450k and EPIC (EPIC V1) array.

Is there a way to use / modify the convertArray function to convert an RGChannelSet of a EPIC V2 array to a RGChannelSet of the EPIC V1 array?

[kunstner]

Same problem here. Do you have any solution to convert EPIC V2 to V1 arrays?

[Pierre9344]

I managed to convert the EPIC v2 id using the sesame R package. There is a liftOver function: https://github.com/zwdzwd/sesame/blob/2d5c2ab371430a8ecb1b5f09792457505a59d192/R/impute.R#L44

[simonjuleseric2]

Could you please share the process?

[kunstner]

The process is implemented in the SeSAMe package. Haven't tried it myself yet but it looks straight forward.

https://doi.org/10.1093/bioinformatics/btae423

I mapped the probes using the annotation files in R. It seems this workflow is suitable for my dataset.

[kostaslazaros]

Hello everyone!

I want to do the same process; convert EPIC V2 to EPIC V1 arrays so that I can use/combine data together at the end.

I have seen the function from the SeSAMe package and it is really interesting. However a point of confusion for me is the data types; Minfi uses RgSet while SeSAMe uses SigDF.

Is there a way to easily convert from one data type to the other?

[kunstner]

I convert the EPICv2 Probe names to EPICv1 after I performed SWAN normalization. I process v1 and v2 arrays separately. Afterwards, I just merge the b and M values.

Assuming you have imported your data using read.metharray.exp you can do something like this

dat <- preprocessRaw(rgSet_v2)
mSetSq_v2 <- preprocessSWAN(rgSet_v2, mSet = dat, verbose = TRUE)

Next, rename the probes

rownames(mSetSq_v2) <- gsub('_BC[0-9][0-9]', '', rownames(mSetSq_v2))
rownames(mSetSq_v2) <- gsub('_TC[0-9][0-9]', '', rownames(mSetSq_v2))

From here, you can extract b/M.

[kostaslazaros]

Many thanks for the quick response!

I am trying to "downgrade" from EPIC V2 to EPIC V2 through the mLiftOver function from Sesame.

My main point of confusion so far is the data type.... is there a way to turn a list of sigDFs into an RGSet ?

[ralanany]

I convert the EPICv2 Probe names to EPICv1 after I performed SWAN normalization. I process v1 and v2 arrays separately. Afterwards, I just merge the b and M values.

Assuming you have imported your data using read.metharray.exp you can do something like this

dat <- preprocessRaw(rgSet_v2) mSetSq_v2 <- preprocessSWAN(rgSet_v2, mSet = dat, verbose = TRUE)

Next, rename the probes

rownames(mSetSq_v2) <- gsub('_BC[0-9][0-9]', '', rownames(mSetSq_v2)) rownames(mSetSq_v2) <- gsub('_TC[0-9][0-9]', '', rownames(mSetSq_v2))

From here, you can extract b/M.

Is it possible to normalize before we combine the data??? Is there reference to do this?

[ralanany]

Thanks for your comment: Is it possible to normalize before we combine the data??? Is there reference to do this?

[kunstner]

Depending on your workflow, you can normalize the data before merging the datasets. For normalization, just follow the steps described in the vignette:

https://bioconductor.org/packages/devel/bioc/vignettes/minfi/inst/doc/minfi.html

[ralanany]

Thanks for your prompt reply, but to my knowledge we have to normalize all the data at the same time to remove the background noise from all samples together. I s there reference or published paper with this work flow?
Thanks again

[kunstner]

We use SWAN normalization on each platform separately before we merge the data. Afterward, it is important to assess the data with respect to potential batch effects. So far, this worked nicely for us but it always depends on the data.