[BioC] PAM: Applying published classifiers

Ryan C. Thompson rct at thompsonclan.org
Fri May 17 22:43:51 CEST 2013


I can't see how the output of pamr.listgenes would be sufficient to 
reproduce a trained classifier. I think your only choice would be to 
re-run PAM starting from the CEL files.

Also, consider whether their classifier would even be applicable to 
your microarray samples, since your samples and theirs are normalized 
separately. If you have a bunch of your own samples that you wish to 
classify, the correct approach might be to normalize the training 
samples and your samples together as one dataset and then re-train the 
classifier, rather than use the exact centroids computed on the 
original normalized data.  In other words, repeating the training 
yourself may be the only statistically valid choice anyway.

On Fri 17 May 2013 01:28:46 PM PDT, Ed Siefker wrote:
> Can someone nudge me in the right direction here?   Am I trying to do
> something
> that isn't possible?  Am I trying to do something that's so obvious it
> hasn't been
> documented?  Am I just unaware of where the appropriate documentation is?
> Any advice would be greatly appreciated.  Thanks
> -Ed
>
>
> On Wed, May 15, 2013 at 1:24 PM, Ed Siefker <ebs15242 at gmail.com> wrote:
>
>> I'm reading through some papers that use PAM to create a classifier from
>> microarray data.
>> I would like to use these classifiers to classify my own samples with
>> microarray data.
>> These papers publish the output of 'pamr.listgenes()', and it's not clear
>> how to massage
>> that into a format that 'pamr.predict()' will accept.
>>
>> The first argument to 'pamr.predict()' is "the result of a call to
>> pamr.train".  'pamr.train()'
>> operates on normalized microarray data and a vector of class labels.
>> Essentially, I'd
>> have to repeat the entire analysis, downloading every CEL file and
>> normalizing it,
>> in order to run 'pamr.train()' so I can run 'pamr.predict'.
>>
>> That doesn't seem like the right way to do things, but I can't find any
>> other function
>> that would create the "pamrtrained" object that 'pamr.predict()'
>> requires.  What's the
>> right way to do what I want to do here?
>>
>
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at r-project.org
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor



More information about the Bioconductor mailing list