[BioC] DESeq2 dispersion estimate gets stuck

[Michael Love]

Just for users who end up on this thread through search: updating to
DESeq2 v1.4.5 clears up this hang.

Mike

On Sat, May 10, 2014 at 8:19 AM, Michael Love
<michaelisaiahlove at gmail.com> wrote:
> hi Carl,
>
> Thanks for reporting this. I had one user report such a hang for
> version 1.2, and had implemented a fix for this in version 1.4, so I'm
> not sure about this one. I could look into this if you could email me
> off list a small example. Is it possible to find a small subset of
> rows which produce the hang?
>
> ddssub <- dds[1:100,]
> ddssub <- estimateDispersions(ddssub)
>
> And then to remove any identifying information,
>
> colData(ddssub) <- colData(ddssub)[, c("patient","proliferation") ]
> levels(ddssub$patient) <- letters[ 1:nlevels(ddssub$patient) ]
>
> Mike
>
> On Fri, May 9, 2014 at 7:51 AM, Carl Herrmann
> <c.herrmann at dkfz-heidelberg.de> wrote:
>> Hi,
>>
>> I am analysing a patient dataset for diff. expression with DESeq2 (v
>> 1.2.10).
>>
>> We have several patients, and for each patient we have several samples.
>> I am running this analysis on different combinations of samples.
>>
>> My design has two factors : patient (multilevel) and a second factor
>> (proliferation, true/false)
>> so : design ~ patient + proliferation
>>
>> I am interested in diff. expression depending on the proliferation only
>> (second factor).
>>
>> On some combinations, the gene dispersion estimation step
>> (estimateDispersion) gets stuck at the gene-wise dispersion estimate.
>> I let it run for several hours, and eventually killed the job.
>>
>> When reducing the number of iterations to 10, it gets stuck. Same problem
>> with version DESeq2 v 1.4.0
>>
>> Strangely, this happens when including samples that, when included in other
>> combinations, run just fine.
>> So this does not seem to be a problem of the dataset itself.
>>
>> I am not sure whether this is enough to get a hint on where the problem
>> might come from, so please tell me whether I should provide additional
>> information.
>>
>>  Thanks for your help !
>>
>> Carl
>>
>> --
>> ------------------------------------------------------------------
>> C a r l    H E R R M A N N
>> ------------------------------------------------------------------
>> Institut für Pharmazie und Molekulare Biotechnologie
>> Universität Heidelberg
>> DKFZ Heidelberg - Department of Theoretical Bioinformatics
>> Im Neuenheimer Feld 580
>> D-69120 Heidelberg
>> tel.: +49 (0) 6221 42-3612
>> email : c.herrmann at dkfz.de
>> web: http://biologie.univ-mrs.fr/carlherrmann
>>
>> _______________________________________________
>> 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