Rma On Probeset Level Means Rma On Exon Level?

Question

[Tanvi]

I am 1st time working on microarray data analysis...My chip is affy human gene 1 st. Using oligo package I have expression values at probeset level (RMA-probeset : around 2.6 lakhs of rows) and at transcipt level(RMA transcript : around 33 thosusands of rows).

My question if transcript level RMA gives gene level expression then what is RMA at probeset level ?

And how it differs with exon array 1 st?

Can I perform exon level analysis using Gene 1 st? It is mentioned on affymetrix site that gene1st is the subset of exon1st and it is very confusing for me.

Thank you in advance

Answer 1

[User 59]

I think the question you're asking is what is the relationship between a probeset and a transcript.

First of all there are multiple probesets to any given transcript. RMA is a summarisation algorithm as well as a normalisation one, so running RMA at the probeset level from the oligo package will return normalised probesets. Running it at the transcript level will return information that is abstracted away from the probeset level into something which more resembled as gene (with splicing variants detectable on an exon array).

As far as I am aware oligo only offers one kind of transcript level summarisation for gene ST arrays, and cannot be used for exon analysis. However oligo's transcript summarisation for exon arrays has more options. I think asking to do an exon analysis without an exon array is perhaps asking a bit much :)

The Gene ST arrays have probes over the entire length of the transcript (about 20) (unlike the original 3' expression arrays), the exon arrays have more probes (about 40) with a minimum number of probes per exon (4). Consequently the exon arrays have greater discriminatory power for splice variant analysis.

[Tanvi]

Thank you Daniel. and gene 1st arrays oligo offers summeries at probeset level. So using probeset level expression data what kind of analysis is possible?

[User 59]

Certainly for the 3' arrays that I analyse most, I analyse probesets not gene level data. Not all probesets are created equal. You can use a probeset as a proxy for a gene. It's just a matter of granularity. I prefer data reported as probeset level data, because it means (if necessary) I can backtrack to which probes provided the measurement. If summarised gene level data is presented, I don't even know what probesets contributed to the data, let alone be able to work out which probes did the measurement. The analysis doesn't differ however.

Answer 2

[Chris Evelo]

As Daniel mentioned you can have multiple probesets for a single gene or gene transcript on a single Affymetrix array. If that is not what you want, or if you just want to have the best probeset annotation you might want to look into custom annotations which are available [?]here[?]. The [?]original publication[?] about customs cdf's appeared in NAR already in 2005.

Answer 3

[David Quigley]

You should read Robinson & Speed BMC Bioinformatics 2007, where they compare the ST and Exon arrays to the older HG133 array. This is a very useful paper, and they specifically consider the question of using the HuGene arrays as exon arrays:

"Like HuEx, HuGene is designed to measure expression over the full length of a transcript. Consequently, there is the possibility that HuGene array could itself be used as an economical exon array for just the well-annotated con- tent. Its exon-level reproducibility is not much worse than the full HuEx array and the coverage of human exons is high."

They don't just say "you should use the ST arrays instead of Exon arrays", but their analysis suggests that this would in fact be a very practical approach and that the probe coverage would be quite similar. The Exon arrays are a lot of work to deal with in my experience.

[User 59]

I thought this was the case but couldn't bring the reference to mind for it though - hardly surprising when it wasn't in my library in the first place. Thanks for the ref!

[Tanvi]

thanks for the ref. It is really helping. I have one more doubt in the file HuGene-1_0-st-v1.na31.hg19.probeset what is exon_id column?

[David Quigley]

Check the README that comes with the file: "Unique identifier for the exon cluster containing this probe set". If that doesn't make sense, read the whitepaper for the array design.