[BioC] Minfi preprocessing and rs probes

[Kasper Daniel Hansen]

On Thu, Nov 22, 2012 at 2:50 AM, Gustavo Fernández Bayón
<gbayon at gmail.com> wrote:
>
> Hi Kasper.
>
> It's perfectly clear for me now what rs* probes are. I should have read the Illumina guide before, I guess. But it is always a pleasure to ask in this list and learn a little bit more.
>
> I agree with your decision about removing the probes. Problem is, GenomeStudio is calculating the betas for those probes too, which is something I think can lead to confusion. In my humble opinion, if something is not measuring methylation, showing a beta value for it on your interface is going to make a new user go mad (as has effectively happened to a fellow biologist in my lab).

Glad that we agree.  If you (or anyone else) have an awesome
application for the rs probes, feel free to tell us.

> I have another question. I was wondering about opening a new topic, but I think it might fit  in this context. Why is minfi not using the correction constant 100 in the denominator when calculating the betas? I have always wondered about that constant, because I did not see the exact reasoning behind it, and I am inclined to think more in your way o implementing it. In the end, it really makes no difference, doesn't it?

Well, they are trying to avoid a divide by zero problem.  Whether or
not that is important to handle also depends on how the data was
normalized.  It is easy to imagine a normalization procedure that
makes sure that this never happens (for example a procedure that
bounds all probe intensities away from zero).  For such a
normalization procedure, adding 100 makes little sense.  In our
implementation, you can choose to add a 100, but that default is zero.
 For most probes this should make no difference, but there may be a
few where it matters.

Kasper


> Regards,
> Gus
>
> ---------------------------
> Enviado con Sparrow (http://www.sparrowmailapp.com/?sig)
>
>
> El miércoles 21 de noviembre de 2012 a las 17:35, Kasper Daniel Hansen escribió:
>
>> The 'rs' probes are normal SNP probes; they do not measure
>> methylation. Quoting from the Illumina methylation guide
>> "rs# representsSNPassays (not affected by DNA methylation)"
>> and
>> "SNP assays can be used for sample identification and tracking. They
>>
>> should be excluded for differential methylation analysis."
>
>>
>> The purpose of these probes is for catching sample mix-ups. We have
>> decided to remove them from the methylation object (MethylSet) because
>> they do not measure methylation and the purpose of this object is to
>> contain the methylation measurements.
>>
>> Kasper
>>
>>
>>
>>
>> On Wed, Nov 21, 2012 at 4:38 AM, Gustavo Fernández Bayón
>> <gbayon at gmail.com (mailto:gbayon at gmail.com)> wrote:
>> > Hi everybody.
>> >
>> > I have just noticed that some probes disappeared during my workflow with minfi. When I looked into the preprocessing, I saw that preprocessIllumina() ends up calling preprocessRaw() in order to generate M and U signals. The interesting part is, only probes of type I and II are used for that. I noticed that the probes that were disappearing in my workflow are of type 'SNPI' and 'SNPII'.
>> >
>> > Now I know in which part the probes are lost, but I would like to know the reason behind this procedure. Also, could anybody give me a hint if this is what Genomestudio is doing? What are this rs* probes useful for then?
>> >
>> > Regards,
>> > Gus
>> >
>> >
>> >
>> > ---------------------------
>> > Enviado con Sparrow (http://www.sparrowmailapp.com/?sig)
>> >
>> > _______________________________________________
>> > Bioconductor mailing list
>> > Bioconductor at r-project.org (mailto:Bioconductor at r-project.org)
>> > https://stat.ethz.ch/mailman/listinfo/bioconductor
>> > Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor
>>
>
>
>