If one gene contains multi-methylation probes, how could I get the methylation level of that gene.

Question

[gongyutang]

I am currently analyzing the TCGA HNSC methylation data. I have split the probes to each gene, but I find out that for one gene, there will be several probes near that gene. I am wondering how could I get the methylation level to represent for that gene if there are mutli_probes near that gene.

I have read TCGA tutorial or other papers, they said that for genes with multi methylation probes, the probe with most anti-correlated is selected to represent the methylation level of a gene. But I am confused that, for methylation, we mostly care about the methylation in the promoter area. Shouldn't I only focus on the probes near the promoter area? So I am thinking about maybe I should get an average value for the probes near the promoter area to represent the methylation level of that gene. Is that Right?

Answer 1

[Kevin Blighe]

It does not make biological sense to obtain the average across the entire gene, unless, perhaps, it's a single exon gene like a non-coding RNA or microRNA. You mention also obtaining the average around the promoter region. This may have greater validity; however, I still think that methylation needs to be treated at the probe-set level. For example, at a bi-directional promoter, one may have a probe on both the 5' and 3'; however, high methylation at just one of these can have consequences to downstream gene transcription. Were you to obtain the average, you may miss things like that if one probe had high methylation and the other not. Even at non-bi-directional promoters, this would be an issue.

Kevin