Problem with "COHCAP.integrate.avg.by.island"

[Anonymous]

Hello,

I'm using COHCAP to correlate my methylation data with the expression data. Everything looks fine when I'm using "by site" workflow but when I want to integrate by island I'm constantly getting the same error suggesting that my expression file does not contain any of my samples. It seems like it writes a warning about all the sample names missing in expression file and then produces NAs ("X9_13 not found in expression file, so all values will be set to 'NA'"), then the NAs are ommited and I get the error "'cor.test.default(methyl, expr, na.action = na.rm)':
'x' must be a numeric vector"
I've checked it hundred of times and it's not true. What can I do? I have even ordered the samples the same way in both data sets.

Any suggestions what can I do about it? I'm not really sure on which step the script assumes that my sample names are missing.

Thanks in advance,
Joanna

[Charles Warden]

Hi Joanna,

I think the issue is that only certain fuctions can be used together:

So, COHCAP.integrate.avg.by.site() has to be used after COHCAP.avg.by.site(), and COHCAP.integrate.avg.by.island() has to be used after COHCAP.avg.by.island().

For the Average-by-Site workflow, the expression file doesn't have any sample names (just a fold-change value, p-value, and FDR value). I've attached copies of the demo expression files for the Bioconductor version. So, if using the Average-by-Site workflow, please see as the example file expression-Average_by_Site_truncated.txt.

In general, the order in the gene expression file shouldn't matter, but I think some users have had an issue with samples that start with numbers (the assumption is either that the Sentrix ID and barcodes for Illumina arrays are used, or sample names that start with a letter are used). In that situation, adding an "S" before the sample names should fix the problem, unless there is actually a sample missing. However, it doesn't sound like that is the problem in this situation.

While I would usually recommend the Average-by-Island workflow over the Averge-by-Site workflow, I think this particular problem can be fixed by using the COHCAP.integrate.avg.by.site() function (and having an appropriately formated expression file).

Best Wishes,
Charles

[Anonymous]

Thank you for your reply, Charles.

Firstly, I have two expression files prepared - one for "by site workflow" and second for the "by island" workflow. I also did follow the necessary steps for both of the workflows. Unfortunately, nothing was really helping with my "COHCAP.integrate.avg.by.island" error that I was constantly getting. I wanted to compare the result from both functions.

But it seems in the meanwhile I've fixed this issue. The problem was following: in all of the other files including sample file and methylation beta table quotation marks are accepted wherever there is a text. I had them also around my sample names (also in the expression file). When I've removed them only from the expression file, everything works perfectly fine. In my opinion the problem was in the perl script that is used inside the function - when it uses temp methylation file without any qoutation marks it can't really locate the same sample names surrounded by them.

Best,
Joanna