[miso-users] Definition of Bayes Factor

[Yarden Katz]

Hi,

First, you're not a member of the mailing list miso-users, so I just noticed that your message was stuck in the queue for approval -- please subscribe if you'd like to receive messages from the list.

The phrase "differential expression" can be applied to a gene or an exon.  In the context of an exon, "differential expression" just means change in Psi value, and this is what the Bayes factor measures.

In your example, the expression level of Exon 2 is relative to Exon 1 and Exon 3 (the flanking exons).  In that case, even if the gene that the exons reside in goes up or down in expression, the Psi value for Exon 2 should not should unless there's a change in splicing, since Exon 2's Psi value is relative to Exon 1 and Exon 3.  

In summary, the Bayes factor measures the change in Psi, and Psi values can change even when there's no change in overall gene expression, or remain constant when there is an overall change in gene expression.  

  --Yarden

On Jan 17, 2013, at 10:39 PM, Ao Zhou wrote:

> Hi, there,
> 
> I'm a new user of MISO.
> I've a question regarding to the definition of BF.
> 
> On the MISO home page, it is said,
> The Bayes factor is computed for the exon, which represents the weight of the evidence in the data in favor of differential expression versus not. 
> For example, a Bayes factor of 2 would mean that the isoform/exon is two times more likely to be differentially expressed than not.
> 
> And later it is said,
> Bayes factor for the comparison, corresponding to the odds of differential expression ,where the change in Ψ is non-zero) over no differential expression (where the change is zero),
> 
> The word "differential expression" is a little confusing to me.
> Let's use an cassette exon event for example:
> 
> [Gene1]
> EXON1---------EXON2------------EXON3
> EXON1--------------------------------EXON3
> 
> Does BF describes the change of EXON2's expression or change in the Psi value?
> 
> As I know the change of EXON2's expression doesn't necessarily comes together with a change of Psi value,
> since its expression is affected by two mechanisms:
> 1) alternative splicing;
> 2) overall expression intensity of Gene1.
> 
> Please correct me if I didn't get it right.
> Thank you!
> -- 
> Ao Zhou
> Ph.D. Student, 
> Bioinformatics Program,
> School of Informatics, IUPUI
> Tel: 317-372-7004
> _______________________________________________
> miso-users mailing list
> miso-users at mit.edu
> http://mailman.mit.edu/mailman/listinfo/miso-users