Hi Lara. > Thanks Mark for your quick answer (this has always amused me, you > always answer so quick...nice)
And getting slower and slower every day :) > so, just some reflexions: > > 1.Because FIRMA is an outlier method, wouldn't it be good to combine > this information with the detection of DE exons and genes in order to > detect alternative splicing events (between conditions)?. This is what > I will do in trying to identify splicing events, with the help of > GenomeGraphs... I'd be interested to hear what you come up with, but some of this is done implicitly in FIRMA(Gene). > 2.It's amazing how variable are genes (even probesets or PSRs) in > different when you look at a probe level (the same gene for the same > sample can have an expression of 2 and 12), so i guess the biology > underneath is much more complicated than our simplifications... I think these are just probe-specific effects, which is the main reason for fitting a model like RMA that builds in probe effects. For example, one of the biggest determinants of raw intensity is the number of Gs/Cs in the probe. There is some biology on top, of course. Mark > > Cheers, > > Lara > > > On 14 Maig, 04:37, Mark Robinson <[email protected]> wrote: >> Hi Lara. >> >> Some comments below. >> >> On 2010-05-14, at 12:48 AM, Lara wrote: >> >> >> >> > Dear all, >> >> > first of all congratulations for aroma, I've been using it for a while >> > and it saved me for the first analysis of exon arrays back in 2008. >> > Since then I have used several times for serveral purposes. >> > Unfortunately, sometimes there is a lack of help but a good forum... >> >> > So, I have been "struggling" with alternative splicing, trying to >> > understand, and done several tests. >> > I have performed Firma scores and limma to see differences between two >> > groups. >> > I have also performed DE exons (probesets) with limma. >> > Then, I used GenomeGraphs to visualize results and I have serveral >> > questions which i hope to write clearly and not to be obvious: >> > 1. what I expect is, in those exons that have differences in fs, to >> > have a clear graphical difference in expression in the selected exon >> > between conditions but is not what I get. >> > So, for instance, if we take 7922737 -- ENSG00000157060 -- C1orf14 -- >> > blue=Testis, which is the first example of supplementary material (1) >> > of "Differential splicing using whole-transcript microarrays", BMC >> > Bioinformatics, 2009, 10, >> 156.http://www.biomedcentral.com/content/supplementary/1471-2105-10-156-s.... >> > I know that is FIRMAGene for gene arrays, but is something similar to >> > what I get with FIRMA and exon arrays. I would say (from the >> > residuals) that last two exons (10 probesets) are spliced. But if you >> > check the expression of those probesets i wouldn't say that they have >> > differences. On the contrary, first exons for instance seem to be >> > spliced. >> > Apparently, DE exons look better on graphs (no matter if they belong >> > to a DE gene or not) >> >> Yes, this can happen. And, that is the real value of the GenomeGraphs >> output. Remember that FIRMA and FIRMAGene are really just outlier >> detection procedures. >> >> This is (sort of) eluded to in both the FIRMA and FIRMAGene papers: >> >> Purdom et al., Section 4.1: "In particular, if the proportion of samples >> showing alternative splicing is high within an exon (say in the majority >> of samples), the high residuals will be found not in those samples >> classified by the simulation as spliced out, but rather the >> complementary set of samples" >> >> Robinson and Speed, Conclusions and Discusion: "Identifying departures >> through residuals from the RMA model will not always be perfect. Some >> departures from the RMA linear model may not be alternative splicing at >> all ... or are induced through, for example, an exon that is not >> expressed in any of the samples in combination with strong differential >> expression." >> >> > 2. Do Firma scores correspond to residuals of rma model? because this >> > is what you plot in genomeGraphs, isn't it? >> >> A FIRMA score for each sample and probeset is the median of the (usually >> 4) residuals from the robustly-fit linear model of RMA. See the paper >> for the formal definition. FIRMAGene is a little bit different, in that >> it takes partial sums of adjacent residuals to try and look for a >> persistence of departure from the model. >> >> > 3. Everything is done on a probeset basis, but shouldn't it be done on >> > a real "exon" (exon cluster) basis?. Sometimes just a probeset appears >> > to be "spliced" and is complicated when you try to interpret this >> > biologically... >> >> Debatable. Affymetrix picked probesets (or "probe selection regions") >> based on what they thought could be independent units of expression, >> based on annotated transcripts, ESTs, etc. But, you could certainly >> build a CDF file that grouped the probes together in a different way. >> >> Hope that helps. >> >> Cheers, >> Mark >> >> > I think there is no point on adding my code and sessionInfo() given >> > that those are "general" questions, but if you need it I can add it. >> >> > Thanks for your time and answer, >> >> > Lara >> >> > -- >> > When reporting problems on aroma.affymetrix, make sure 1) to run the >> latest version of the package, 2) to report the output of >> sessionInfo() and traceback(), and 3) to post a complete code example. >> >> > You received this message because you are subscribed to the Google >> Groups "aroma.affymetrix" group with >> websitehttp://www.aroma-project.org/. >> > To post to this group, send email to [email protected] >> > To unsubscribe and other options, go >> tohttp://www.aroma-project.org/forum/ >> >> ------------------------------ >> Mark Robinson, PhD (Melb) >> Epigenetics Laboratory, Garvan >> Bioinformatics Division, WEHI >> e: [email protected] >> e: [email protected] >> p: +61 (0)3 9345 2628 >> f: +61 (0)3 9347 0852 >> ------------------------------ >> >> ______________________________________________________________________ >> The information in this email is confidential and intended solely for >> the addressee. >> You must not disclose, forward, print or use it without the permission >> of the sender. >> ______________________________________________________________________ >> >> -- >> When reporting problems on aroma.affymetrix, make sure 1) to run the >> latest version of the package, 2) to report the output of sessionInfo() >> and traceback(), and 3) to post a complete code example. >> >> You received this message because you are subscribed to the Google >> Groups "aroma.affymetrix" group with >> websitehttp://www.aroma-project.org/. >> To post to this group, send email to [email protected] >> To unsubscribe and other options, go >> tohttp://www.aroma-project.org/forum/ > > -- > When reporting problems on aroma.affymetrix, make sure 1) to run the > latest version of the package, 2) to report the output of sessionInfo() > and traceback(), and 3) to post a complete code example. > > > You received this message because you are subscribed to the Google Groups > "aroma.affymetrix" group with website http://www.aroma-project.org/. > To post to this group, send email to [email protected] > To unsubscribe and other options, go to > http://www.aroma-project.org/forum/ > ______________________________________________________________________ The information in this email is confidential and intended solely for the addressee. You must not disclose, forward, print or use it without the permission of the sender. ______________________________________________________________________ -- When reporting problems on aroma.affymetrix, make sure 1) to run the latest version of the package, 2) to report the output of sessionInfo() and traceback(), and 3) to post a complete code example. You received this message because you are subscribed to the Google Groups "aroma.affymetrix" group with website http://www.aroma-project.org/. To post to this group, send email to [email protected] To unsubscribe and other options, go to http://www.aroma-project.org/forum/
