MAJIQ voila doubts

[Swethaa NG]

Hi,

I came across some specific questions to ask about voila. Sorry for a bunch of questions, I hope they are clear enough.

Background: I had some mice samples to check for differential splicing: 3 controls and 2 mutated samples. (group 1 : Controls and group 2 : mutated)

These samples were run in the same batch. I used heterogen quantifier since I thought it would be more specific to the individual samples. Then, I ran $voila modulize and $voila tsv to obtain the results

1. How different are modulize results from voila tsv?

 I want to clarify whether modulize is a simple representation of the individual and summarized events strictly within the stats thresholds, and if voila tsv is a list of all the modules with prob_changing (false and true) events? 

($voila tsv splicegraph.sql /file/C-Homozygous.het.voila -f /file/C-Homozygous-voila.tsv ... and I did not mention -showall. But, I could still see all the non significant modules listed)

Modulize summary showed only one gene module, and voila tsv had the same gene module being the only 'prob_changing' significant one. So I am guessing the results would always be the same in both?

2. When I checked voila view, I could see more genes listed there than compared to voila  modulize. They seem to be similar to what is seen in voila tsv. The default settings on the browser remained unchanged. Please guide me on how to interpret this?

3. In voila view, for one module, the dPSI value for a LSV is -0.59. May I please know where to view this value in the voila tsv file? I do not seem to see a matching value..

4. Is it okay that I kept the controls as group 1? How much do the results differ based on the group labels?

Thanks a lot! :D

Appreciate your time!

Kind regards,

Swethaa

[San Jewell]

Hi Swethaa,

There are good questions here. Perhaps we could work to make some things clearer in the official documentation as for the differences between the various tools available. For now I will answer your questions as best I can with the information you provided:

Voila view differences: Voila view is an exploratory tool intended to assist with a visual representation of genes of interest, finding genes of interest, and generating figures. There are some operations which are compute heavy, such as calculating stats with a different confidence threshold for every lsv in the entire dataset, in order to filter the index table. It would currently be unusable to recalculate these values exactly for filtering the table in a reasonable live usage scenario. Instead, the confidence values are binned into 10 bins and saved in an index for swift but less exact filtering in voila view, which may lead to more/less lsvs being shown for the same filters compared to in other modes. Voila tsv and voila modulize will always calculate the exact thresholds.

Voila tsv vs voila modulize: Voila TSV basically takes the output of quantifier (het, psi, dpsi) and combines it with information from the splicegraph to generate a more complete human readable output. At the same time, a wide range of filters are provided to help narrow the result to information of interest. Meanwhile, modulizer has many of the same filters, but it's goal is much different. It can take any number of quantifier files (which will become output columns in the final result) and it begins by making modules from the filtered splicegraph, Then, inside these modules, it will look for a number of pre-defined splicing patterns and sort them into various output files. (Mode information on the splicing patterns here: https://biociphers.bitbucket.io/majiq-docs/modulizer/event-types.html)

For either tsv mode or modulize mode, if you supply DPSI or HET quantifier files, the default behavior is to output only _changing_ Lsvs, which will eliminate the majority of the data, most of the time. You can see all of the data instead by supplying the --show-all switch. If you supple a het file, and still see something which is not changing, there is either a value you missed somewhere which would make the program believe it was changing, or there is a bug of some sort, and we can look further into this specific case.

dpsi value in voila not matching value in TSV -- Can you please show the relevant screenshot or voila and the rows of the tsv file you are comparing? I just want to make sure we are on the same page.

The group labels do not really change any outcome ; of course please make them all unique to each other.

Let me know if it's understandable,

-San