[miso-users] Running MISO on replicates

[Yarden Katz]

Hi,

There's no built-in support for this; see below for more information from mailing list archives.  In short, you have to post-process the Psi values/Bayes factors for samples within the same condition/group, and if standard analyses (clustering, etc.) show that groups behave similarly, you can also consider pooling the samples for some analyses to gain power (e.g. merging the BAM files and running MISO on the aggregate BAM file.)

Best, --Yarden

--
>From miso-users-owner at PCH.MIT.EDU Thu Apr 19 21:55:38 2012

Hi Jerry,

If I understand correctly, you have 4 replicates for the control condition, and 4 for the KO condition.  The question is then how does MISO deal with replicates.  The short answer is that MISO right now only performs pairwise comparisons -- so in this case, you could compare each control with each KO sample, and vice versa.  However, once you generate the summary files and MISO output for each sample, there are several other ways to deal with replicates in general, and the most appropriate one will depend on what downstream analyses you're doing. 

Take a look at Question 13 from the FAQ page in the MISO manual -- it contains some information on this:

http://genes.mit.edu/burgelab/miso/docs/#frequently-asked-questions-faq

Best, --Yarden





On Aug 16, 2013, at 5:29 PM, Manjusha Pande wrote:

> 
> Hi Yarden,
> 
> I am running MISO on two groups of 3 samples each ((WT1, WT2, WT3, Mutant1, Mutant2 and Mutant3). I ran the steps up to summarize-samples on the individual samples. I want to run the differential expression step on the two groups.
> 
> Under FAQ, I see a description on how to run MISO on replicates. I was wondering if you could elaborate on how to pool samples within a group for downstream analysis. An example of run_miso.py command would be very helpful.
> 
> Thank you.
> 
> MP
> _______________________________________________
> miso-users mailing list
> miso-users at mit.edu
> http://mailman.mit.edu/mailman/listinfo/miso-users