LSV splicing type annotation

[Oliver Ziff]

Dear developers,

I am trying to make biological sense of the voila tsv output. I am interested in global whole genome splicing changes in the cannonical splicing types between conditions (ES, MXE, IR, A5SS, A3SS etc). The voila tsv does not annotate each LSV with its splicing type and so I am doing this myself. The vast majority of events are complex with >2 Num. Junctions but i wondered how to unpick their composition and look for regular patterns. I wondered if you had any improvements on this:

 

voila_tsv <- voila_tsv %>%
      mutate(splicing_type = case_when(`Num. Junctions` == 2 & ES == "False" & A5SS == FALSE & A3SS == FALSE ~ "retained intron",

                         `IR coords` != "" ~ "complex: retained intron",

                         `Num. Junctions` == 2 & `Num. Exons` == 3 ~ "skipped exon",

                         `Num. Junctions` == 3 & `Num. Exons` == 4 ~ "two-skipped exons",

                         `Num. Junctions` == 4 & `Num. Exons` == 5 ~ "multi-skipped exons",

                         `Num. Junctions` == 5 & `Num. Exons` == 6 ~ "multi-skipped exons",

                         `Num. Junctions` == 6 & `Num. Exons` == 7 ~ "multi-skipped exons",

                         `Num. Junctions` == 7 & `Num. Exons` == 8 ~ "multi-skipped exons",

                         `Num. Junctions` == 8 & `Num. Exons` == 9 ~ "multi-skipped exons",

                         `Num. Junctions` == 9 & `Num. Exons` == 10 ~ "multi-skipped exons",

                         `Num. Junctions` == 2 & `Num. Exons` == 4 ~ "mutually exclusive exon",

                         `Num. Junctions` == 2 & A3SS == TRUE ~ "alternative 3\'",

                         `Num. Junctions` == 2 & A5SS == TRUE ~ "alternative 5\'",

                         TRUE ~ "complex: other"))

Many thanks,
Oliver

[Christian Rohde]

Hi Oliver and dear MAJIQ developers,

this is actually a great question. I also would like to have a global overview about the splice changes between conditions. It would be nice if one could directly export a global view figure from voila view using the parameters chosen. On the other hand I also wish to understand how I can work with the information and merge with expression changes etc. I started working with MAJIQ some days ago and it seems that the information MAJIQ provides changed recently. I do not find the cols ES, A5SS, A3SS. What I get is num_junctions, num_exons and ir_coords. In addition I have the col lsv_type which is not self explanatory for my. I do not find a hint in the help how to deal with information like :

• s|1e1.1o1|i
• s|1e1.1o1|1e2.1o1 
• s|1e1.1o1|1e2.1o1 
• t|1e1.1o1|1e2.1o1
  

Is there a way to translate the lsv_type into human readable format?

Best,

Christian

[Christian Rohde]

sorry for spamming. I found a tsv from in my dpsi output folder from DeltaPSI calculation which includes all LSV and the columns you mentioned. Just the file I exported with voila tsv for downstream analyses as suggested in the help is different. At this point the help is a bit confusing ...

On Wednesday, July 1, 2020 at 11:13:12 PM UTC+2 Oliver Ziff wrote:

[Denisa Bogdan]

Hi christian,

Did you ever find an answer for this question? I am hoping to understand the lsv_type column and I am really confused.

Many thanks,

Denisa

[Paul Jewell]

Hi All, 

I'm a little unsure of what the goal is in this case. The purpose of the "lsv_type" representation is only for the ease of displaying the cartoon in voila view, so, for the best visual representation, I would highly recommend using voila view to check the lsv_types, however, for a rough translation:

• starts by (s or t) being source or target
  
• each '|' is a new junction representation and if there is intron_retention the last character is 'i'
• each junction is represented by  XeY.ZoK where 
  • X is the ordinal splice site in the reference exon
  • Y is the ordinal exon connecting the lsv
  • Z is the ordinal splice site in exon Y
  • K is the total number of splice sites that Y has

Again, though, it is in house code, not generally stable, and not meant to be parsed. 

Thanks, 

-Paul

[iamnic...@gmail.com]

Hi Majiq devs,

I want to echo Oliver and Christian's interest in some kind of biological interpretation of the lsv_type column. My specific goals are to compare the usage of different categories of AS events (binary, complex, etc) between groups globally. I am also of course using deltapsi + Voila to look at gene-by-gene differences, but also something similar to what was reported in figures 4 and 5 of the 2016 eLife paper. Maybe this is already explained somewhere in the docs and i've missed it!

Thanks!

Nico

[Matthew Gazzara]

Hi Nico,

In our new updates to MAJIQ / Voila we just published (https://www.nature.com/articles/s41467-023-36585-y) we introduced the Voila Modulizer (https://biociphers.bitbucket.io/majiq/VOILA_modulize.html) to do what you and others have been asking for. This will produce a number of text files that each correspond to the various types of alternative splicing events. A summary file is also generated to show which alternative splicing modules exist in the splice graph (a single entry, single exit region) and which event types they are made up of. Check out Figure 4 for how this works. 

On your data you can use the decomplexify threshold options to remove junctions with low PSI and/or low dPSI values before you define the AS modules or AS events. Doing this will make the output AS events and AS modules less complex and easier to get a meaningful count from. 

All the best,

Matt

[Matthew Gazzara]

Additional Modulizer documentation is also available here https://biociphers.bitbucket.io/majiq-docs/modulizer/index.html

[iamnic...@gmail.com]

Hi I know this isn't the place, but I'm trying to get a build of Majiq that has the modulizer feature. I never got a confirmation email from the download form (filled out last week) and emailed the biocipher admins shortly thereafter. Apologies for being such a nag! Just really wanna try out this software!

Nico

[San Jewell]

Hi Nico,

I've replied to you from other channels. However, the takeaway that I posted here is that for some reason it seems the majiq email is being spam-filtered my a number of email providers by default. I would recommend for anyone else seeing this issue to please check your spam folder while we check why this might be happening.

Thanks,

-San

[iamnic...@gmail.com]

Hi San,

That's very weird. This google group email certainly did end up in junk but that is the only correspondence from the Biociphers I've found in any of my inboxes, junk, and deleted mail folders. I'll try to email you again from my .edu email as I am currently running the 2.4.dev3+g427c4efe Majiq build and need an update.

Thanks and sorry for this pain in the neck,

Nico

[San Jewell]

Hi Niko,

Did you receive the other email from me (sje...@pennmedicine.upenn.edu) with the setup instructions? I sent it manually.