[miso-users] new mouse genome annotation for alternative cleavage and polyadenylation events

[Yarden Katz]

On Jan 7, 2013, at 5:35 PM, Charizanis,Konstantinos wrote:

> Thank you Yarden for the fast reply!
> This is the error it reports:
> 
> Traceback (most recent call last):
>  File "/apps/python/2.7.3/bin/filter_events.py", line 406, in <module>
>    main()
>  File "/apps/python/2.7.3/bin/filter_events.py", line 403, in main
>    apply_both_samples=options.apply_both)
>  File "/apps/python/2.7.3/bin/filter_events.py", line 104, in multi_filter
>    apply_both_samples=False)
>  File "/apps/python/2.7.3/bin/filter_events.py", line 265, in filter_events
>    for bf in bayes_factor:
> TypeError: 'float' object is not iterable

It's a cryptic error and on my list to change it, but yes, "filter_events.py" is only intended for two isoform events and will fail if you feed it a comparison file where there are entries with more than 2 isoforms.

It doesn't make sense to filter on a single-Bayes factor threshold in multi-isoform mode, since you have multiple isoforms to worry about.  2 out of the 5 isoforms might be differential, or just 1, or 3 etc. so it's left to the user to decide how to parse that.

> 
> 
> I understand what you say. To clarify something, does this mean that the 5 different values for each tab (for example diff: -0.00,0.00,-0.03,-0.24,0.27) is the value for each one of the new annotated 3'UTR? 
> I already performed the analysis with the older annotation (two events per gene) but I got excited to get a more complete view of the polyA changes with the new annotation.
> I guess I won't be able to use it since I cannot filter out the results, unless I break it down in different columns and try to run it that way (probably I'm wrong). 

The diff vector "-0.00,0.00,-0.03,-0.24,0.27" is a number saying what the difference in Psi is for each UTR isoform in your pair of samples.  So the first two UTR isoforms have a DeltaPsi of 0, while the third isoform changes by -3% and the fourth and fifth change by -24% and -27%, respectively.

Many people use multiple isoform annotations instead of the two isoform annotation (not just for 3' UTRs, but for all RNA processing events.)  It's more complex to work with, but MISO outputs all the information you need.  Filtering is not that simple since there isn't a single number for each event/gene, but a vector of numbers.  I.e. a vector of delta Psi values, a vector of Bayes factors, and so on, instead of a single number in the two-isoform case.  Since there are many ways to filter these results, MISO does not have a built in utility for doing that, it's up to you to decide what kind of changes are meaningful in your analyses when there are several annotated UTRs.

Best, --Yarden

> 
> Thank you again!
> 
> 
> -----Original Message-----
> From: Yarden Katz [mailto:yardenny at gmail.com] On Behalf Of Yarden Katz
> Sent: Monday, January 07, 2013 4:20 PM
> To: Charizanis,Konstantinos
> Cc: miso-users at mit.edu
> Subject: Re: [miso-users] new mouse genome annotation for alternative cleavage and polyadenylation events
> 
> Hi Konstantinos,
> 
> See comments below:
> 
> On Jan 7, 2013, at 3:31 PM, Charizanis,Konstantinos wrote:
> 
>> I used the new annotation (Bin Tian DEC 2012)  to do a previous analysis but I have the following problem. Miso works up to the comparing sample step. After that it gives an error when I try to filter the results.
>> I went and viewed the Bayes factors file output of the comparison and it doesn't make any sense. I'm copied the first couple rows.
>> 
>> 
>> event_name      sample1_posterior_mean  sample1_ci_low  sample1_ci_high sample2_posterior_mean  sample2_ci_low  sample2_ci_high diff    bayes_factor   etc.
>> 
>> Zyx:42307851    0.01,0.03,0.05,0.48,0.44        0.00,0.01,0.01,0.15,0.02        0.01,0.05,0.08,0.90,0.77        0.01,0.03,0.08,0.73,0.16        0.00,0.01,0.04,0.44,0.01        0.01,0.04,0.11,0.89,0.45    0.00,0.00,-0.03,-0.24,0.27      0.00,0.03,0.13,0.95,0.93        'Zyx:42307851.1.0','Zyx:42307851.2.0','Zyx:42307851.3.0','Zyx:42307851.4.0','Zyx:42307851.5.0'
> 
> Note that the new Tian annotation is a *multiple isoforms* annotation, meaning it annotates two or more 3' UTRs per gene.  The output you show above is the correct format for a gene with multiple (in this case, five) isoforms.  You get a Psi value for each isoform, a DeltaPsi for each isoform, etc. 
> 
> You never said what the error you got was -- but the built-in filtering feature in MISO is only defined for two-isoform events, which is probably why it did not work.
> 
> You could use the new Tian annotation and modify your analyses to account for the fact that some genes have more than 2 isoforms.  Otherwise, you could use the old annotation, or select only genes with two isoforms from the new Tian annotation if those are the genes that interest you.
> 
>> 
>> 
>> I also checked the indexed gff annotation and it looks different from the previous annotations that exist on the MISO website.  
> 
> Yes, it's a totally different annotation, and the older one is still available.  
> 
> Best, --Yarden
> 
>> Any input or ideas would be much appreciated.
>> Thank you
>> 
>> Konstantinos
>> _______________________________________________
>> miso-users mailing list
>> miso-users at mit.edu
>> http://mailman.mit.edu/mailman/listinfo/miso-users
>