[BioC] Beta score of DNA methylation assays

[Sean Davis]

On Mon, Jun 1, 2009 at 12:04 PM, Adaikalavan Ramasamy
<a.ramasamy at imperial.ac.uk> wrote:
> Sonia Shah, thanks for the link. The Beta score is calculated as you said
> with the addition that any negative values of Cy3 and Cy5 are set to zero.
>
> Sean Davis, thanks for pointing out to the methylumi package. It looks like
> a nice package. I am slightly concerned that you are analyzing the Beta
> scores without any data transformation (though I like how you illustrated it
> using the limma package). The Beta values appear to be highly bimodal and
> the model near zero seems to be more pronounced. Having said that, I also
> seen others using t-test to analyze the data too so I think data
> tranformation may be a more general problem. Or maybe I am worrying too
> much.

Hi, Adai.

Thanks for the comments. The goal of the package is really to parse
the data and apply a reasonable normalization to the data.  There are
certainly robust tests that could be used for statistical purposes.
That said, the distribution of Beta values within a sample across CpG
sites (which always has the bimodality you mention) is not the
distribution of interest, I'm sure you know better than I (I'm no
statistician!).  Instead, you want to examine the distribution of beta
values for a given CpG site within the two populations of interest.  I
have not looked at this in any detail, but if those two distributions
approach normality, then a t-test may be appropriate.  So, in short,
there is a bit of a gray area here.  I should probably add some
disclaimer to the vignette to that effect.

Sean


>
> Saroj Mohaptra, thanks for pointing out to Houseman et al. I don't fully
> understand this method yet but will look further into it. Thanks.
>
> Thank you all.
>
> Regards, Adai
>
>
>
> Sonia Shah wrote:
>>
>> Hi Adai,
>>
>> I think the beta score is the ratio of intensities between methylated and
>> unmethylated alleles:
>>
>> Beta = methylated signal/(unmethylated signal + methylated signal + 100)
>>
>> So in your case Beta= Cy5/(Cy3+Cy5+100).
>>
>> A Beta of 0 means the site is completely unmethylated and a value of 1
>> means
>> fully methylated.
>>
>> See:
>> http://www.illumina.com/downloads/DNAMethylationAnalysis_DataSheet.pdf
>>
>>
>>
>> Regards,
>> Sonia
>>
>>
>>
>>
>> Adaikalavan Ramasamy wrote:
>>>
>>> Hi there,
>>>
>>> I been handed a tabular file from DNA methylation containing "Beta"
>>> values (rows are CpG island and columns are subjects) from Illumina
>>> platform. The Beta values range between 0 and 1.
>>>
>>> I am wondering how this "Beta" value is derived, what it means and how to
>>> use it.
>>>
>>> Or do I calculate the Cy3/Cy5 ratio and proceed as with gene expression
>>> microarray studies?
>>>
>>> Please accept my apologies if these are basic questions and kindly point
>>> me to any reference. It is my first DNA methylation dataset. Thank you.
>>>
>>> Regards, Adai
>>>
>>> _______________________________________________
>>> Bioconductor mailing list
>>> Bioconductor at stat.math.ethz.ch
>>> https://stat.ethz.ch/mailman/listinfo/bioconductor
>>> Search the archives:
>>> http://news.gmane.org/gmane.science.biology.informatics.conductor
>>>
>>
>>
>>
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives:
> http://news.gmane.org/gmane.science.biology.informatics.conductor
>