[BioC] flowClust results in subset filters
Florian Hahne
fhahne at fhcrc.org
Wed May 6 01:36:42 CEST 2009
Hi Kenneth,
did you also update your flowClust? I had to add the summarizeFilter
method there. The version you need is 2.3.1.
When running the last command of the example code I get this
names(filterDetails(fr, "tmixFilter in defaultRectangleGate"))
[1] "filter" "populations" "expName" "varNames"
"K"
[6] "w" "mu" "sigma" "lambda"
"nu"
[11] "z" "u" "label" "uncertainty"
"ruleOutliers"
[16] "flagOutliers" "rm.min" "rm.max" "logLike"
"BIC"
[21] "ICL"
Those are all the additional slots from the tmixFilterResult class.
Florian
Kenneth Lo wrote:
> Hi Florian,
>
> Thanks a lot for the fix, which clearly makes the integration of
> flowClust with flowCore more sense as now it may tell the class
> assignment from tmixFilter when combining its use with other
> filtering/gating operations defined in flowCore. I've just checked
> out the most updated copy from the devel branch (version 1.11.2), and
> tried the example code in your email. The filter method in the 2nd
> last line of the example code does return a multipleFilterResult
> object now, which is what we desire. However, similar to Bastian's
> experience, I could only access the tmixFilter object but not the
> tmixFilterResult object in the filterDetails list. Some slots which
> describe the fitted distribution using flowClust, e.g., mu, sigma and
> lambda, are found in the tmixFilterResult class but not in
> tmixFilter. Are there any means to retrieve these pieces of information?
>
> Thanks again,
>
> Kenneth
>
>
>
>
> On 5-May-09, at 11:41 AM, Florian Hahne wrote:
>
>> Hi Bastian, Kenneth,
>> the problem in this case is that flowClust extends the
>> multipleFilterResult class, adding all the mixture parameters and
>> stuff. The subsetFilter operation is defined in flowCore, which
>> doesn't really know about all these additional slots. We do have a
>> mechanism in place to deal with such cases, i.e, the summarizeFilter
>> method which is called whenever a filter is evaluated. This is
>> basically just a hook that has access to both the input filter and
>> the filterResult, and it can be used to add arbitrary items to the
>> filterDetails slot. I added a summarizeFilter method for tmixFilters
>> which does exactly that.
>> The following will still give you a multipleFilterResult object as
>> the result of the subsetFilter operation but now with additional
>> flowClust parameters attached in the filterDetails.
>>
>> library(flowClust)
>> data(GvHD)
>> foo <- GvHD[[1]]
>> rg <- rectangleGate("FSC-H"=c(100,600), "SSC-H"=c(200, 700))
>> tm <- tmixFilter(parameters=c("FSC-H", "SSC-H"), K=3)
>> fr <- filter(foo, tm %subset% rg)
>> names(filterDetails(fr, "tmixFilter in defaultRectangleGate"))
>>
>> Note that there are three items in the filterDetails list, one for
>> each of the components of the subsetFilter and one for the
>> subsetFilter itself, and that the latter contains copies of values of
>> the former two. This is intentional, albeit a bit unfortunate in this
>> case, since the additional flowClust parameters contain a lot of
>> information (large matrices, indexing vectors...), which is
>> unnecessarily copied here.
>>
>> bw,
>> Florian
>>
>>
>> Bastian Angermann wrote:
>>> Thank you very much and have a great weekend.
>>>
>>> Bastian
>>>
>>> -------- Original-Nachricht --------
>>>
>>>> Datum: Fri, 01 May 2009 14:16:01 -0700
>>>> Von: Florian Hahne <fhahne at fhcrc.org>
>>>> An: Bastian Angermann <AngerB at gmx.de>
>>>> Betreff: Re: [BioC] flowClust results in subset filters
>>>>
>>>
>>>
>>>> Ok, I will take a look asap,
>>>> Florian
>>>>
>>>> Bastian Angermann wrote:
>>>>
>>>>> Hi Florian,
>>>>>
>>>>> I've come over a minor issue with the current return type of a
>>>>> filter operation that subsets a tmix-filter within an other filter.
>>>>>
>>>>> While the current return type of multipleFilterResult allows to
>>>>> access
>>>>> the class assignment by the tmix-Filter, the results describing
>>>>> the fitted distribution (i.e. means, proportions and covariance
>>>>> matrices) are not accessible. The filter included in the
>>>>>
>>>> filterDetails in the appropriate gate is of type tmixFilter and not
>>>> tmixFilterResult, which would of course have a redundant subSet.
>>>>> Access to the parameters of the fitted distributions would be
>>>>> quite usefull.
>>>>>
>>>>> Best,
>>>>> Bastian
>>>>> -------- Original-Nachricht --------
>>>>>
>>>>>> Datum: Mon, 27 Apr 2009 17:41:43 -0700
>>>>>> Von: Florian Hahne <fhahne at fhcrc.org>
>>>>>> An: Bastian Angermann <AngerB at gmx.de>
>>>>>> Betreff: Re: [BioC] flowClust results in subset filters
>>>>>>
>>>>>
>>>>>> Hi Bastian,
>>>>>> I just committed the changes to the devel branch of the package
>>>>>>
>>>> (version
>>>>>> 1.11.1).
>>>>>> It might take a day or two to build, but it would be great if you
>>>>>> could
>>>>>> take a look at this version first. I would like to hold off on
>>>>>> bumping things back into release unless we are all happy with it.
>>>>>> If you know how to build the devel package from source you should
>>>>>> be able to get the 1.11.1 version as of now.
>>>>>> Florian
>>>>>>
>>>>>> Bastian Angermann wrote:
>>>>>>
>>>>>>> Hi Florian,
>>>>>>>
>>>>>>> thanks for your reply. Let me know if I can do any beta testing
>>>>>>> of the new code.
>>>>>>>
>>>>>>> Cheers,
>>>>>>> Bastian.
>>>>>>> -------- Original-Nachricht --------
>>>>>>>
>>>>>>>> Datum: Sun, 26 Apr 2009 22:49:02 -0700
>>>>>>>> Von: Florian Hahne <fhahne at fhcrc.org>
>>>>>>>> An: Bastian Angermann <AngerB at gmx.de>
>>>>>>>> CC: Bioconductor at stat.math.ethz.ch
>>>>>>>> Betreff: Re: [BioC] flowClust results in subset filters
>>>>>>>>
>>>>>>>
>>>>>>>> Hi Bastian,
>>>>>>>> this is indeed a feature which wasn't implemented yet. I fixed
>>>>>>>> the code and everything seems to be working as intended now,
>>>>>>>> but I want
>>>>>>>>
>>>> to
>>>>>>>> give it some more rigorous testing before submitting. I am
>>>>>>>> planning
>>>>>>>>
>>>> to
>>>>>>>> push that back into the release branch since it is more of a
>>>>>>>> bug fix than feature addition.
>>>>>>>> If you want to be more independent of the types of comparisons
>>>>>>>>
>>>> between
>>>>>>>> gating steps, you might want to take a look at the workFlow
>>>>>>>> stuff (as
>>>>>>>>
>>>>
>>>>>>>> described in the package vignette). There you can compute
>>>>>>>> summary statistics for a particular gating operation based on
>>>>>>>> another reference view in the workFlow (via the additional
>>>>>>>> 'reference' argument of the summary method)
>>>>>>>> bw,
>>>>>>>> Florian
>>>>>>>>
>>>>>>>>
>>>>>>>> On 24.04.2009, at 10:30, Bastian Angermann wrote:
>>>>>>>>
>>>>>>>>
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> I am using R2.9.0 and the 2.4 version of Bioconductor. When I
>>>>>>>>> try to
>>>>>>>>>
>>>>
>>>>>>>>> run
>>>>>>>>> a tmixfilter subsetted on a rectangular gate the result is of
>>>>>>>>> class logicalFilterResult thus I am unable to access the
>>>>>>>>> individual clusters:
>>>>>>>>> #################################################
>>>>>>>>> library(flowCore)
>>>>>>>>> library(flowClust)
>>>>>>>>> data(rituximab)
>>>>>>>>>
>>>>>>>>> r_filter <- rectangleGate("FSC.H"=c(100,500),"SSC.H"=c(100,500))
>>>>>>>>> # gate boundaries are not meant to be meaningful.
>>>>>>>>> t_filter <- tmixFilter("tmix", c("FSC.H", "SSC.H"), K=3)
>>>>>>>>> result <- filter(rituximab,t_filter %subset% r_filter)
>>>>>>>>>
>>>>>>>>> class(result)
>>>>>>>>> ################################################
>>>>>>>>>
>>>>>>>>> Is that a bug, missing feature or am I missing something?
>>>>>>>>> I'd like to be able to see the result of the
>>>>>>>>> tmixFilter in the context of the ancestor populations,
>>>>>>>>> thus the %subset%. I'am working around that by adding an index
>>>>>>>>>
>>>> vector
>>>>
>>>>>>>>> to the original flowFrame, to be able to do manual subsetting,
>>>>>>>>> but
>>>>>>>>> it would be nice to to that without this hack.
>>>>>>>>>
>>>>>>>>> Thanks a lot,
>>>>>>>>> Bastian Angermann
>>>>>>>>> --
>>>>>>>>>
>>>>>>>>> _______________________________________________
>>>>>>>>> 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
>>>>>>>>
>>>>>>>
>>>>>> --
>>>>>> Florian Hahne, PhD
>>>>>> Computational Biology Program
>>>>>> Division of Public Health Sciences
>>>>>> Fred Hutchinson Cancer Research Center
>>>>>> 1100 Fairview Ave. N, M1-B514
>>>>>> PO Box 19024
>>>>>> Seattle, Washington 98109-1024
>>>>>> 206-667-3148
>>>>>> fhahne at fhcrc.org
>>>>>>
>>>>>
>>>> --
>>>> Florian Hahne, PhD
>>>> Computational Biology Program
>>>> Division of Public Health Sciences
>>>> Fred Hutchinson Cancer Research Center
>>>> 1100 Fairview Ave. N, M1-B514
>>>> PO Box 19024
>>>> Seattle, Washington 98109-1024
>>>> 206-667-3148
>>>> fhahne at fhcrc.org
>>>>
>>>
>>>
>>
>>
>> --
>> Florian Hahne, PhD
>> Computational Biology Program
>> Division of Public Health Sciences
>> Fred Hutchinson Cancer Research Center
>> 1100 Fairview Ave. N, M1-B514
>> PO Box 19024
>> Seattle, Washington 98109-1024
>> 206-667-3148
>> fhahne at fhcrc.org
>>
>
--
Florian Hahne, PhD
Computational Biology Program
Division of Public Health Sciences
Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N, M1-B514
PO Box 19024
Seattle, Washington 98109-1024
206-667-3148
fhahne at fhcrc.org
More information about the Bioconductor
mailing list