Pierre beat me to this one. Comments below...
On Thu, Dec 5, 2013 at 9:20 AM, Pierre Neuvial
<pierre.neuv...@genopole.cnrs.fr> wrote:
> Hi Emilie,
>
> OK, so you are referring to the “smooth.CNA" function in the DNAcopy
> package, cf
> http://www.bioconductor.org/packages/2.13/bioc/vignettes/DNAcopy/inst/doc/DNAcopy.pdf
>
> What this function is doing is detecting outliers (based on how far their
> signal value is from their neighbors) and shrink their signal values toward
> those of their neighbors.
>
> This is indeed appropriate and recommended. I thought that by "smoothing"
> you meant performing some kind of local averaging of the original signal
> (e.g. using a mobile median or by binning): this I don't recommend. Sorry
> for the confusion.
>
>
> To drop outliers, one possibility is to use the "dropSegmentationOutliers"
> function from the PSCBS package. See the vignettes at
> http://cran.fhcrc.org/web/packages/PSCBS/index.html
>
> Another comment: since you are following the vignette for paired CNA
> analysis, I am guessing that you are working with tumor/normal pairs. If
> so, then you should use PSCBS rather than CBS for segmentation. PSCBS is an
> extension of CBS to segment not only total copy numbers but also allelic
> ratios. See the PSCBS vignette in the above URL.
To balance this a little bit, I would say there may exist outliers in
the total copy number (TCN) signals that are so sever that they bias
the estimators/test statistic of CBS (which assumes Gaussian signals).
If one believes there are such outliers and worries that they are so
extreme that they would affect the segmentation severely, one could
either (i) drop or (ii) shrink ("smooth") them. In the vignettes of
the PSCBS package, I've last night [PSCBS (>= 0.39.8)]
corrected/clarified Section 'Dropping TCN outliers' to say the
following:
"There may be some outliers among the TCNs. In
CBS~\citep{OlshenA_etal_2004,VenkatramanOlshen_2007}, the authors
propose a method for identifying outliers and then to shrink such
values toward their neighbors ("smooth") before performing
segmentation. At the time CBS was developed it made sense to not just
to drop outliers because the resolution was low and every datapoint
was valuable. With modern technologies the resolution is much higher
and we can afford dropping such outliers, which can be done by:
> data <- dropSegmentationOutliers(data)
Dropping TCN outliers is optional."
Hope this clarifies.
Back to the original question: It is not possible to drop (or smooth)
outliers using the CbsModel() pipeline [I'll add that to the todo
list]. The easiest is to turn use the PSCBS package, where you can do
plain old single-track CBS segmentation, paired PSCBS segmentation and
also non-paired PSCBS segmentation. As Pierre says, if you have tumor
SNP data, you should look into doing parent-specific CN analysis,
which you can do either via paired or non-paired PSCBS depending on
whether you have match normals or not.
To take your allele-specific CRMAv2 and bring it into a format
recognized by the PSCBS package, see
http://aroma-project.org/vignettes/PairedPSCBS-lowlevel
/Henrik
>
> Best,
>
> Pierre
>
>
> On Wed, Dec 4, 2013 at 5:29 PM, Emilie <emilie.lalo...@gmail.com> wrote:
>>
>> Hi Pierre,
>>
>> Thanks for your answer. I may be wrong but I thought smoothing prior to
>> segmentation was somewhat common. It is shown in the vignettes for DNACopy
>> and seems to be fairly common in the literature (this approach was used in
>> the Metabric paper for example,
>> http://www.ncbi.nlm.nih.gov/pubmed/22522925).
>>
>> I'd be interested in hearing more of your thoughts against this. Do you
>> have an idea of how much resolution is lost by smoothing?
>>
>> Emilie
>>
>>
>>
>> On Tuesday, December 3, 2013 5:26:38 PM UTC-5, Pierre Neuvial wrote:
>>>
>>> Hi Emilie,
>>>
>>> It's certainly possible to do this within the Aroma framework (e.g. using
>>> the function "binnedSmoothing"). It's probably not as straightforward as
>>> running the segmentation directly, though, because this is not a typical use
>>> case.
>>>
>>> In fact, I'm not sure why you want to perform smoothing before
>>> segmentation ? Smoothing is definitely not required before segmentation,
>>> and I would actually discourage to go this path because it will end up in a
>>> loss of resolution along the genome at the smoothing step.
>>>
>>> Best,
>>>
>>> Pierre
>>>
>>>
>>> On Tue, Dec 3, 2013 at 8:53 PM, Emilie <emilie....@gmail.com> wrote:
>>>>
>>>> Hi there,
>>>>
>>>> I'm new to processing Affy SNP6 chips and so am mainly experimenting
>>>> with different methods to date. I ran CRMAv2 and followed steps 1-4 from
>>>> the
>>>> vignette (http://aroma-project.org/vignettes/CRMAv2). For step 5, I want to
>>>> do a paired analysis.
>>>>
>>>> Previously I've used DNAcopy to perform CBS for other array types, and
>>>> would like to follow a similar procedure, which includes smoothing prior to
>>>> segmentation. Is this possible using the aroma.affymetrix package? So far
>>>> I've followed the vignette for paired CNA analysis
>>>> (http://aroma-project.org/vignettes/pairedTotalCopyNumberAnalysis) but
>>>> haven't seen any options for smoothing.
>>>>
>>>> thank you very much,
>>>>
>>>> emilie
>>>>
>>>> --
>>>> --
>>>> When reporting problems on aroma.affymetrix, make sure 1) to run the
>>>> latest version of the package, 2) to report the output of sessionInfo() and
>>>> traceback(), and 3) to post a complete code example.
>>>>
>>>>
>>>> You received this message because you are subscribed to the Google
>>>> Groups "aroma.affymetrix" group with website http://www.aroma-project.org/.
>>>> To post to this group, send email to aroma-af...@googlegroups.com
>>>>
>>>> To unsubscribe and other options, go to
>>>> http://www.aroma-project.org/forum/
>>>>
>>>> ---
>>>> You received this message because you are subscribed to the Google
>>>> Groups "aroma.affymetrix" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send
>>>> an email to aroma-affymetr...@googlegroups.com.
>>>>
>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>
>>>
>> --
>> --
>> When reporting problems on aroma.affymetrix, make sure 1) to run the
>> latest version of the package, 2) to report the output of sessionInfo() and
>> traceback(), and 3) to post a complete code example.
>>
>>
>> You received this message because you are subscribed to the Google Groups
>> "aroma.affymetrix" group with website http://www.aroma-project.org/.
>> To post to this group, send email to aroma-affymetrix@googlegroups.com
>> To unsubscribe and other options, go to
>> http://www.aroma-project.org/forum/
>>
>> ---
>> You received this message because you are subscribed to the Google Groups
>> "aroma.affymetrix" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to aroma-affymetrix+unsubscr...@googlegroups.com.
>> For more options, visit https://groups.google.com/groups/opt_out.
>
>
> --
> --
> When reporting problems on aroma.affymetrix, make sure 1) to run the latest
> version of the package, 2) to report the output of sessionInfo() and
> traceback(), and 3) to post a complete code example.
>
>
> You received this message because you are subscribed to the Google Groups
> "aroma.affymetrix" group with website http://www.aroma-project.org/.
> To post to this group, send email to aroma-affymetrix@googlegroups.com
> To unsubscribe and other options, go to http://www.aroma-project.org/forum/
>
> ---
> You received this message because you are subscribed to the Google Groups
> "aroma.affymetrix" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to aroma-affymetrix+unsubscr...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
--
--
When reporting problems on aroma.affymetrix, make sure 1) to run the latest
version of the package, 2) to report the output of sessionInfo() and
traceback(), and 3) to post a complete code example.
You received this message because you are subscribed to the Google Groups
"aroma.affymetrix" group with website http://www.aroma-project.org/.
To post to this group, send email to aroma-affymetrix@googlegroups.com
To unsubscribe and other options, go to http://www.aroma-project.org/forum/
---
You received this message because you are subscribed to the Google Groups
"aroma.affymetrix" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to aroma-affymetrix+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
