[BioC] tilingArray- normalizeByReference

[Anjan Purkayastha]

Anjan Purkayastha wrote:
> joern,
> thanks.
> let me modify a local version and see the results. is it ok if i work 
> on this problem with you? i think dr. huber's method to genomic dna 
> hybridization based approach suits our purpose the best.
> to give you a brief idea about our platform: we have built a whole 
> genome tiling array for the vaccinia virus (ds DNA virus, genome size: 
> ca 190kb); there are about 19000 overlapping probes, 60 bp in length, 
> for each strand. for the purposes of normalization i have removed 
> overlapping probes and am left with 6488 non-overlapping, contiguous 
> members.
>
> anyway, i'll rerun the procedure with the revised script and see what 
> i come up with .
> thank you for your help.
> warm regards,
> anjan
>
>
>
>
> Joern Toedling wrote:
>> Anjan,
>> looking at the source code of the function normalizeByReference, there
>> is indeed the line
>> vsnMatrix(xn, lts.quantile = 0.95, subsample = as.integer(2e+05)) ,
>> and since your matrix only has 6488 lines, this cannot work.
>> Without knowing more details about your array platform and study, I
>> cannot say whether or not the function "normalizeByReference" is really
>> applicable to your array platform. However, we should augment the
>> function anyway by chaning this line to something like
>> vsnMatrix(xn, lts.quantile = 0.95, subsample = min(nrow(xn),
>> as.integer(2e+05))) .
>> We are going to make this change and it will be included in the
>> development version of tilingArray within a few days. For the moment,
>> you will have to work with a thus modified local copy of the function.
>> Sorry for the inconvenience.
>>
>> Regards,
>> Joern
>>
>>
>> Anjan Purkayastha wrote:
>>  
>>> joern,
>>> thanks. i've solved the problem by wrapping an as. integer function
>>> around the pm_vector. so here is my command:
>>>
>>> vac_normalized_data= normalizeByReference(vac_exp_set, vac_hyb_set,
>>> pm= as.integer(pm_vector), background= as.integer(bg_vector))
>>>
>>> however my story does not end here. i get a further set of errors
>>> after this command.
>>> Error in validObject(.Object) :
>>>  invalid class "vsnInput" object: 'subsample' must be a numeric vector
>>> of length 1 with values between 0 and 6488.
>>> In addition: Warning message:
>>> In normalizeByReference(vac_exp_set, vac_hyb_set, pm =
>>> as.integer(pm_vector),  :
>>>  'some strata of background contain fewer than 5000 features, are you
>>> sure this is alright?
>>> Error in exprs(vsnMatrix(xn, lts.quantile = 0.95, subsample =
>>> as.integer(2e+05),  :
>>>  error in evaluating the argument 'object' in selecting a method for
>>> function 'exprs'
>>>
>>> the first part of the error message: vsnInput" object: 'subsample'
>>> must be a numeric vector of length 1 with values between 0 and 6488,
>>> does not make much sense to me.
>>> if i understand the second part of the error message about background
>>> strata correctly, there are less than 5000 features in some background
>>> strata. that is understandable as i have only 1275 probes in the
>>> background set. there are a total of 6488 probes in my set.
>>>
>>> any ideas on how to troubleshoot this?
>>> thanks.
>>> anjan
>>>
>>>
>>>
>>>     
>>
>>   
>
>


-- 
===========================================
anjan purkayastha, phd
bioinformatics analyst
whitehead institute for biomedical research
nine cambridge center
cambridge, ma 02142

purkayas [at] wi [dot] mit [dot] edu  
703.740.6939