Question: Problem when mapping non-core exons with MTA1.0 arrays
0
4.0 years ago by
jacorvar40
European Union
jacorvar40 wrote:

Dear bioC community,

I have an ExpressionSet object called eset, whose annotation is mta10stprobeset.db. When I try to map the probesets to Entrez IDs  as mta10stprobesetENTREZID(featureNames(eset)), I get the error:

Error in .checkKeys(value, Lkeys(x), x@ifnotfound) :    value for "TC0100000001.mm.1" not found

Could you tell me why this happens and what I could do to solve it?

Thanks

modified 4.0 years ago by James W. MacDonald52k • written 4.0 years ago by jacorvar40
Answer: Problem when mapping non-core exons with MTA1.0 arrays
0
4.0 years ago by
United States
James W. MacDonald52k wrote:

TC0100000001.mm.1 is a transcript level probeset, so you should be using the mta10sttranscriptcluster.db package to annotate.

In other words, the probeset you are talking about is essentially the same set of probes regardless of whether you summarize at the probeset or core level. However, at the probeset level it should be PSR0100000001.mm.1, and at the core level it should be TC0100000001.mm.1.

> dbGetQuery(con, "select * from featureSet where man_fsetid='PSR0100000001.mm.1';")
fsetid         man_fsetid strand   start    stop transcript_cluster_id
1 20846209 PSR0100000001.mm.1      0 3054233 3054733     TC0100000001.mm.1

Here man_fsetid is the ID at the probeset level, and the transcript_cluster_id is the ID at the core level. Since you have at least one ID from the core level, it implies that you ran rma() using the default of summarizing at the core level, and hence you should be using the mta10sttranscriptcluster.db package to annotate.

This is me inferring things that may not actually be true, based on what limited information you have given us. The lesson for you is that you shouldn't make it so people have to guess as to what you have done, and instead you should be providing more information, including the code you used to summarize. The lesson for me (if in fact you have summarized at the probeset level, and this is actually a bug) is that I shouldn't be inferring things and instead should be telling people to give more information so we don't have to make blind inferences.