[miso-users] Differentially expressed isoform

[Yarden Katz]

Hi Rob,

See replies below:

On Mar 7, 2012, at 8:22 AM, Li, Robert wrote:

> Hello everyone,
> I am a new user of MISO and have 2 small questions.
> I attempt to detect differentially expressed isoforms between 2 samples (a pairwise comparison).
> I notice that for a gene with 2 isoforms, MISO output calculated only one value for Bayes factor between 2 samples (2 isoforms). However, for a gene with 3 isoforms, the Bayes factor was calculated for each of 3 isoforms (see attached).

For two isoforms, the BF and the Psi value are reported as single numbers for convenience.  The Bayes factor corresponds to the odds ratio in favor of differential regulation of the first isoform in the two-isoform case.  Naturally, if the first changes, and there are only two isoforms, this implies that the second one changes as well.  Similarly, in the two isoform case, the Psi value is the expression of the first isoform in the annotation, and the expression of the second is simply 1 - (first Psi value).

In the general case of more than two isoforms, the Psi values and Bayes factors are reported for each isoform.  So if you have three isoforms A,B,C, the Psi vector will be the expression level of each of those, and the Bayes factor will correspond to the degree to which each isoform changes across the samples.  If there are only 3 isoforms A,B,C, then the Psi values and BFs of these will be dependent, since their expression level of A,B,C must sum to 1.  

> In addition, some of my Bayes factor is as large as 3.60E+140. Is this common?

Yes.  Just as in hypothesis testing with p-values, you can sometimes get arbitrarily high (or in the case of p-values, arbitrarily low) values.  I recommend seeing which BF thresholds make sense in your data empirically, e.g. see how the quantity of interest (e.g. % of differential events regulated) changes between your samples as you increase the BF threshold.  See MISO paper for discussion and example of this. 

The BF is meant to serve as a "significance threshold" not as a truly quantitative measure.  So if you use a cut off of BF >= 20, I don't think an event with a BF of 10,000 is any less "significant" than an event with BF of 500,000.

Best, --Yarden

> Thank you very much!
> Rob
> This electronic message contains information generated by the USDA solely for the intended recipients. Any unauthorized interception of this message or the use or disclosure of the information it contains may violate the law and subject the violator to civil or criminal penalties. If you believe you have received this message in error, please notify the sender and delete the email immediately. <MISO_Sample_C1_vs_MISO_Sample_BT4.miso_bf.filtered>_______________________________________________
> miso-users mailing list
> miso-users at mit.edu
> http://mailman.mit.edu/mailman/listinfo/miso-users