450k methylation array - 2 raw intensity values for same probe

Question

[neuro3030]

From GEO omnibus I have download raw intensities from a 450k methylation array (in a tab-delimited text format). However, there are duplicates of some of the probes:

example for Sample 1:

There are two rows for this probe, one for Address A and one for Address B:

1. cg00000622 (Address = 11642304)
2. cg00000622 (Address = 38691301)

However, each has a methylated and an unmethylated intensity

1. methylated = 907 unmethylated = 10835
2. methylated = 120 unmethylated = 735

How do I get this down to single methylated and unmethylated values for this probe? (so that I may calculate the beta value).

Is this a type I probe or is one of them a control probe?

Thanks for any help

[ATpoint]

This is a two-color array. You should get a background towards that platform first before diving into analysis, e.g. https://www.bioconductor.org/help/course-materials/2015/BioC2015/methylation450k.html

[neuro3030]

Thank you. But how does one arrive at a single intensity value for methylated and a single intensity value for unmethylated, for type I probes? For type II probes, the methylated is in the red channel, unmethylated in green. For type I, am I to interpret the red ("methylated") channel as having both a methylated and unmethylated signals? It's a bit confusing.

If red and green do not equate to methylated and unmethylated for type I probes, how does one calculate the beta value from this? Do you add both methylated signals = M, and add both unmethylated signals = U?

Answer 1

[Charles Warden]

I wouldn't normally calculate this myself.

If you have access to the .idat files, you can use GenomeStudio and/or minfi to test various normalizations when calculating beta values.

In GEO, I thought you were required to provide both intensity and beta values, but that does make testing alternative processing more difficult.

[neuro3030]

Thanks. However, I do have a specific problem:

The raw intensity values I downloaded from GEO contain the above format. All raw intensities I have downloaded from GEO from the 450k platform thus far contain the usual 485,577 probes. However, this particular data-set contains 617,984 probe intensities, due to listing of both intensity values for type I probes: methylated and unmethylated values for each red and green channels.

This causes an error of "duplicate" rows when import the data into minfi. How do I fix this?

Thanks

[Charles Warden]

I apologize for the delayed response.

I changed the formatting to try and make myself a little more clear.

Time-permitting, I will see if I can understand what you are describing better by looking at another data set.

So, I am not 100% certain what to suggest, but I am surprised that beta values were not already provided (and I think the best solution is if you can re-process the raw .idat files).

[Charles Warden]

Maybe there is something that I am not taking into consideration, but I just checked the intensity values for this dataset, and there are 485,577 rows.

ftp://ftp.ncbi.nlm.nih.gov/geo/series/GSE42nnn/GSE42308/suppl/GSE42308_HY_450k_signal_intensity.txt.gz

So, I am not sure if there are multiple valid ways to submit intensity value to GEO, but that is different than the data that I have uploaded.