[BioC] error in hugene10sttranscriptcluster

[James W. MacDonald]

Hi Javier,

On 4/18/2011 1:27 PM, Javier Pérez Florido wrote:
> Hi Jim,
> See below
>
> On 18/04/2011 16:00, James W. MacDonald wrote:
>> Hi Javier,
>>
>> On 4/18/2011 7:02 AM, Javier Pérez Florido wrote:
>>> Dear Jim,
>>> Thanks for your quick reply.
>>> I'm not sure if I understood your explanation. I did some research at
>>> the netaffx annotation as you suggested.
>>>
>>> I've summarized using oligo package at the "core" level, i.e. transcript
>>> level. I've checked through NetAffx annotation files all the transcript
>>> clusters related to the NR_002716 gene.
>>> These are: 7948894, 8019631,8019633,8019635,8019637,8019639, 8019641,
>>> 8019703,8019705,8019707,8019709 and 8019802. What does this mean? A gene
>>> is made up of several clusters? Is the gene repeated through these
>>> clusters? Once I understand this issue, I will understand how limma
>>> works on these arrays, since I don't know whether cluster=gene or
>>> clusterS = gene.
>>
>> This is one of the problems inherent to the design of arrays that use
>> really short probes (and often very few probes per probeset). Since
>> the probes are so short, they sometimes will bind to many different
>> transcripts, so it is hard to say which transcript is responsible for
>> the signal for a given probeset.
>>
>> If you put 8019631 in netaffx, you will see that the hybridization
>> target is 'mixed', meaning it binds multiple transcripts. The only
>> probeset that comes up if you query netaffx for NR_002716 is 7948894,
>> so I assume all the others you note are mixed as well.
> So, if I understood well, many transcripts may hybridize to the same
> probeset so that, when the hybridization is unique, the gene/transcript
> intended to hybridize is shown by the annotation package and when many
> transcripts may hybridize to the probeset, a NA is given because we
> don't really know which gene is hybridized to the probeset, is that
> correct? At least, in the annotation packages developed by the Biocore
> data team (this does not happen in hugene10sttranscriptcluster.db or in
> NetAffx biological annotation given by Oligo)
>
> My only concern is related to 7948894 probeset. When it is introduced in
> netaffx, it says that this probeset is mixed (the transcripts detected
> by this probeset are NR_002716, NR_002761, AK292330 and X59360). I'm not
> a biologist, so, at first, I would say that, if this probeset is
> "detected" by limma as being differentially expressed, we can not
> distinguish which gene or transcript is really differentially expressed
> since there are four possible (unless they are related in some way).

Exactly. There is no way to tell how much of the signal is due to a 
particular transcript. But since you aren't a biologist, note that each 
gene can produce many different transcripts due to splice variants. And 
what we are talking about here are transcripts, not genes.

Also note that accession numbers point to transcripts, not genes. And 
the accession numbers are assigned when researchers define what they 
think are new transcripts (but may not actually be new).

So there can be multiple accession numbers per gene that may or may not 
be different transcripts of that gene. For instance, both NR_002716 and 
X59360 point to the same gene, and NR_002761 points to a closely related 
gene. I don't know offhand if NR_002716 and X59360 are different 
transcripts (e.g., splice variants) or not. But you may need to.

Best,

Jim


>
> It's true that when NR_002716 is introduced in netaffx, it says that the
> probeset related is 7948894, but it seems that this probeset can detect
> also NR_002761 and others.
>
> Am I right?
> Thanks again, this discussion is being very useful for me,
> Javier
>
>>
>>>
>>> I observed some differences between the annotated files
>>> HuGene-1_0-st_v1.na29.hg18.transcript (the one I used before) and
>>> HuGene-1_0-st_v1.na31.hg19.transcript (the latest one). The main
>>> differences are related to the "start" and "stop" fields on these files
>>> for each of the transcript clusters described above.
>>> For the first version (na29.hg18), there are numbers different from zero
>>> on these fields, whereas in the latest version (na31.hg19), the "start"
>>> and "stop" values are zero. However, in both files, the "gene
>>> assignment" field is NR_002716. So, I don't understand why when I use
>>> mget("8019631",hugene10sttranscriptclusterACCNUM) and error is found
>>> whereas in the NetAffx annotation file this accession number exists.
>>
>> For the annotation packages we supply (we meaning the Biocore Data
>> Team), any probesets that interrogate multiple transcripts are set to
>> return NA by default. You can change this behavior using the
>> toggleProbes() function. I did a quick test of that the other day, and
>> it doesn't appear that was done with this package. But you might try
>> yourself.
>>
>> You might also contact the maintainer directly (we don't make these
>> packages), to see if he can give more complete answers.
>>
>> Best,
>>
>> Jim
>>
>>
>>>
>>> Moreover, when using the annotation from oligo (which retrieves NetAffx
>>> Biological Annotation):
>>> pData(featureData(OligoEset))["8019631","geneassignment"]
>>> returns NR_002716
>>>
>>> I'm a little bit confused about this.
>>> Thanks again,
>>> Javier
>>>
>>>
>>> On 16/04/2011 20:40, James MacDonald wrote:
>>>> Hi Javier,
>>>>
>>>> The annotation of Affy chips tends to change over time, and this might
>>>> be an instance of that. If you check netaffx for this probeset, the
>>>> transcript it measures is described as 'multiple', and if you blat the
>>>> sequence they built the probeset against, it matches all over the
>>>> place.
>>>> So it may be that in the past they claimed a direct match and now they
>>>> don't.
>>>>
>>>> You could investigate this further by looking at older versions of the
>>>> annotation files if you care to know more.
>>>>
>>>> Best,
>>>>
>>>> Jim
>>>>
>>>>
>>>>
>>>> James W. MacDonald, M.S.
>>>> Biostatistician
>>>> Douglas Lab
>>>> 5912 Buhl
>>>> 1241 E. Catherine St.
>>>> Ann Arbor MI 48109-5618
>>>> 734-615-7826
>>>>>>> Javier Pérez Florido 04/16/11 8:16 AM>>>
>>>> Dear list,
>>>> I'm trying to get the ENTREZIDs of some Affy_IDs of GeneChip Human Gene
>>>> ST 1.0 Arrays through hugene10sttranscriptcluster package. Depending on
>>>> the R version, the results are different.
>>>> For example, in R 2.12.2:
>>>> > mget("8104901",hugene10sttranscriptclusterENTREZID)
>>>> $`8104901`
>>>> [1] "3575"
>>>>
>>>> But
>>>> >mget("8019631",hugene10sttranscriptclusterENTREZID)
>>>> Error en .checkKeys(value, Lkeys(x), x at ifnotfound) :
>>>> value for "8019631" not found
>>>>
>>>> The sessionInfo is:
>>>>
>>>> sessionInfo()
>>>> R version 2.12.2 (2011-02-25)
>>>> Platform: x86_64-pc-mingw32/x64 (64-bit)
>>>>
>>>> locale:
>>>> [1] LC_COLLATE=Spanish_Spain.1252 LC_CTYPE=Spanish_Spain.1252
>>>> [3] LC_MONETARY=Spanish_Spain.1252 LC_NUMERIC=C
>>>> [5] LC_TIME=Spanish_Spain.1252
>>>>
>>>> attached base packages:
>>>> [1] grid tools tcltk stats graphics grDevices utils
>>>> [8] datasets methods base
>>>>
>>>> other attached packages:
>>>> [1] annotate_1.28.1 oneChannelGUI_1.16.5
>>>> [3] girafe_1.2.0 genomeIntervals_1.6.0
>>>> [5] intervals_0.13.3 ShortRead_1.8.2
>>>> [7] lattice_0.19-17 Rsamtools_1.2.3
>>>> [9] Biostrings_2.18.4 GenomicRanges_1.2.3
>>>> [11] baySeq_1.4.0 edgeR_2.0.5
>>>> [13] IRanges_1.8.9 preprocessCore_1.12.0
>>>> [15] GOstats_2.16.0 graph_1.28.0
>>>> [17] Category_2.16.1 tkWidgets_1.28.0
>>>> [19] DynDoc_1.28.0 widgetTools_1.28.0
>>>> [21] affylmGUI_1.24.0 affyio_1.18.0
>>>> [23] affy_1.28.0 limma_3.6.9
>>>> [25] hugene10sttranscriptcluster.db_6.0.1 org.Hs.eg.db_2.4.6
>>>> [27] RSQLite_0.9-4 DBI_0.2-5
>>>> [29] AnnotationDbi_1.12.0 Biobase_2.10.0
>>>>
>>>> loaded via a namespace (and not attached):
>>>> [1] BSgenome_1.18.3 genefilter_1.32.0 GO.db_2.4.5
>>>> GSEABase_1.12.2
>>>> [5] hwriter_1.3 RBGL_1.26.0 splines_2.12.2
>>>> survival_2.36-5
>>>> [9] XML_3.2-0.2 xtable_1.5-6
>>>>
>>>> However, in R 2.10.0
>>>> mget("8104901",hugene10sttranscriptclusterENTREZID)
>>>> $`8104901`
>>>> [1] "3575" (the same as before in R 2.12.2)
>>>>
>>>> > mget("8019631",hugene10sttranscriptclusterENTREZID)
>>>> $`8019631`
>>>> [1] "6066" (there is no error like in R 2.12.2)
>>>>
>>>> The sessionInfo is:
>>>> R version 2.10.0 (2009-10-26)
>>>> i386-pc-mingw32
>>>>
>>>> locale:
>>>> [1] LC_COLLATE=Spanish_Spain.1252 LC_CTYPE=Spanish_Spain.1252
>>>> [3] LC_MONETARY=Spanish_Spain.1252 LC_NUMERIC=C
>>>> [5] LC_TIME=Spanish_Spain.1252
>>>>
>>>> attached base packages:
>>>> [1] stats graphics grDevices utils datasets methods base
>>>>
>>>> other attached packages:
>>>> [1] limma_3.2.3
>>>> hugene10sttranscriptcluster.db_4.0.1
>>>> [3] org.Hs.eg.db_2.3.6 RSQLite_0.9-2
>>>> [5] DBI_0.2-5 AnnotationDbi_1.8.2
>>>> [7] Biobase_2.6.1
>>>>
>>>> loaded via a namespace (and not attached):
>>>> [1] tools_2.10.0
>>>>
>>>> Why this error for Affy_ID 8019631 when R2.12.2 is used?
>>>> Thanks,
>>>> Javier
>>>>
>>>> _______________________________________________
>>>> 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
>>>>
>>>> **********************************************************
>>>> Electronic Mail is not secure, may not be read every day, and should
>>>> not be used for urgent or sensitive issues
>>>>
>>>>
>>>
>>
>

-- 
James W. MacDonald, M.S.
Biostatistician
Douglas Lab
University of Michigan
Department of Human Genetics
5912 Buhl
1241 E. Catherine St.
Ann Arbor MI 48109-5618
734-615-7826
**********************************************************
Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues