Converting beta value to M value (when beta=1)

Question

[janhuang.cn]

There are actually two questions.

1) I am using the ComBat function in sva package to adjust for batch effect (plate). Before adjusting, the beta values range from 0.0016 to 0.9997. After adjusting, the beta values range from -0.4152 to 1.4015. What is the recommended way to deal with this? Should I just arbitrarily make all beta values smaller than 0 to 0, and all beta values larger than 1 to 1?

2) I am using the beta2m function in lumi package to convert beta values to M values. But when the beta value equal to 1, NaN is returned. Should I make all 1 to (1-0.00001) or something like this? But should I use 1-0.00001 or 1-0.001 or something else? Is there a guideline?

It would be best if anyone could also provide reference.

Thank you.

Answer 1

[Devon Ryan]

1. Why not use ComBat with the M values?
2. Either keep values some minimum distance from 0/1 or convert NaNs to sufficiently large numbers (I've used -13 and 13 in the past I think, since that was around the limit of machine precision).

[janhuang.cn]

Thank you. I now converted beta to m-value, and then do combat with m-value, and then convert the combat-adjusted m-value to combat-adjusted beta. This solve the problem

[wuhan9281]

Hi professor. I am using the IlluminaHumanMethylationEPICanno.ilm10b4.hg19 Bioconductor package to get the annotation for some CpG sites. But I found that UCSC_RefGene_Name for some CpGs has more than one gene.

There are two scenarios.

1) The same gene is listed multiple times. For example,

UCSC_RefGene_Name of cg09315878 is "SDF4;SDF4". This is splice variants. I wonder if I could interpret this as cg09315878 is on SDF4 gene?

2) Different genes are listed for one CpG. For example,

UCSC_RefGene_Name of cg02716826 is "SUGT1P1;AQP3". I am not sure why and how to interpret this. And then for GO enrichment analysis, which gene name should I used?

[Devon Ryan]

You're replying to an almost 4 year old post.

1. Yes, if it overlaps multiple transcripts then by definition it overlaps the gene.
2. Welcome to micoarrays, probes aren't always specific.