[BioC] DESeq - fitNbinomGLMs - Error: NA/NaN/Inf in foreign function call

[Wolfgang Huber]

Dear Ioannis

thank you for the feedback! This should not happen.

Would you be able to send us:
- the output of "sessionInfo()"
and, if at all possible,
- the R code to reproduce your problem and a copy of cdsFull (perhaps 
with anonymized gene names if that's a consideration for you).

One observation is that your size factors vary a lot (and this seems 
broadly to align with type a or b). If the data quality is very 
different for batches of the experiment, fitting the "type" effect in 
the GLM might not be enough to compensate for all the batch-related 
effects in your data, and perhaps separately normalising and testing the 
"a" and "b" samples, and then combining / overlapping the resulting gene 
lists or p-values might be more advisable ("meta-analysis").

	Best wishes
	Wolfgang


Il Jun/28/11 5:59 AM, ALok ha scritto:
> Hi Ioannis
>
> Perhaps there is a 0 value, which is passed inside the function, needed to
> compute some internal functions. It may explains lack of convergence in
> glm.fit function.
> I hope that this helps,
>
> Alok
>
> On Tue, Jun 28, 2011 at 12:18 AM, Filippis, Ioannis<
> i.filippis at imperial.ac.uk>  wrote:
>
>> Hi,
>>
>> I am trying to run fitNbinomGLMs but depending on the factor selected, it
>> gives error "Error: NA/NaN/Inf in foreign function call (arg 1)".
>> The counts matrix doesn't have any Na/NaN values, I cannot really
>> understand what the problem is. I guess it shouldn't work for any factor if
>> there was any problem with gene counts.
>>
>>> pData(cdsFull)
>>      sizeFactor condition type
>> wtH1 2.17904027   treated    a
>> wtP1 0.24260614   treated    b
>> prH1 1.42343596 untreated    a
>> prP1 0.14120460 untreated    b
>> wtH4 3.45554434   treated    a
>> wtP4 2.38048956   treated    b
>> prH4 4.13642641 untreated    a
>> prP4 0.40209227 untreated    b
>> wtH7 3.76575182   treated    a
>> wtP7 0.88389937   treated    b
>> prH7 2.98917685 untreated    a
>> prP7 0.08601356 untreated    b
>>> head(counts(cdsFull))
>>   wtH1 wtP1 prH1 prP1 wtH4 wtP4 prH4 prP4 wtH7 wtP7 prH7 prP7
>> 1   12    3    5    0   12   30   10    0    8   11   11    0
>> 2  195   14   99    0   14   31   38    2   36   12   39    1
>> 3  135   25  112   15  482  252  321   34  233   73  175    4
>> 4  149   11  119   22  525  181  828   92  998  181 1086   19
>> 5    6    0    8    0    5    3    3    0    2    0   12    0
>> 6   39    2    4   21   77  120   26  192   35   15    6   11
>>
>>> fit0<-fitNbinomGLMs(cdsFull, count ~ type)
>> .............
>> There were 26 warnings (use warnings() to see them)
>>> fit1<-fitNbinomGLMs(cdsFull, count ~ condition)
>> ..........Error: NA/NaN/Inf in foreign function call (arg 1)
>> In addition: There were 21 warnings (use warnings() to see them)
>>> fit1<-fitNbinomGLMs(cdsFull, count ~ condition + type)
>> Error: NA/NaN/Inf in foreign function call (arg 1)
>> In addition: Warning message:
>> glm.fit: algorithm did not converge
>>
>> Many thanks for your help.
>>
>> Best,
>> Ioannis
>> _______________________________________________
>> 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
>>
>
>
>


-- 


Wolfgang Huber
EMBL
http://www.embl.de/research/units/genome_biology/huber