Mismatch b/w Majiq Voila TSV and browser view...

[Chris Khoury]

Hi Majiq Team!

 

Thanks for your great splicing pipeline (MAJIQ v2.4).

 

We have been looking at using "voila tsv" to replicate results on the command line, based on the majiq voila browser (v2.4) for a given CI and expected threshold.

 

The objective is to save what we see in the voila browser (plus all of its column data) by using the tsv argument of voila.

 

In the browser we input the abs(E(dPSI)) Threshold and Confidence Threshold filters.

 

Using Voila tsv at the command line, we have set various floats using the 2 listed parameters under "Thresholds for deltapsi inputs" :

 

--threshold (e.g 0.20, 0.19, 0.15)

 

--probability-threshold (e.g 0.95).

 

However, we cannot reproduce in the tsv output the list of LSVs that we see in the browser.

 

Specifically, we retrieve 2 LSVs in the browser, and 35, when using the tsv filters at the command line.

 

Do you know what might be happening here?

 

Thanks in advance for the pointers,

Chris.

[San Jewell]

Hello Chris,

You are correct that currently the filters applies in the view and the tsv are different. This may be amended in a future release, and only applies to classic dPSI mode, not het or psi.

In the TSV mode, things are simple, the probability of the event above *threshold* must be >= *probability threshold*

In view mode it's a bit more complex. This is because we do not on the fly calculate the probability based on *threshold* for every event (when threshold cannot be indexed because it can change by user) This would slow down the interactive index page too much. The compromise is that we bin into 10 possible thresholds (0.0, 0.1, 0.2, etc), and the slider just snaps to the actual threshold behind the scenes. This is why it will not be exactly the same. The index page should generally show you events of interest, but for the most accurate one shot results the TSV mode is better.

Let me know if it makes sense.

-San

[Chris Khoury]

Hi San, Happy new year!

Thanks for your note.

Could you briefly expand upon the binning process you have mentioned in relation to the attached screenshots of our data?

The first screenshot is the visualisation from Voila (majiq_pic_1).

The second is a screenshot of the tsv at the same settings with a changing threshold of 0.2 and CI of 0.95 (majiq_pic_2).

The gene names are blacked out in both screenshots but if we take Gene A and B as examples would you be able to advise in that context of where the other 33 no-named genes that met the 0.20 requirement got binned to?

Thanks in advance,
Chris

[Chris Khoury]

Hello San, 

I hope you are well today.

Please let me know if my question makes sense however could you briefly expand upon my note on Jan 13th regarding the binning process in relation to the 2 screenshots I posted; one from from Voila and the other being the corresponding TSV file.

Thanks very much.

Appreciate the help,

Chris 

[San Jewell]

Hi Chris,

Can you let me know what form of information you would like in order to best help with your query? For example, I can not actually step through the calculations and figure out where the threshold is being passed/not passed without the data. If you would like I can try to provide some excerpts of the relevant code which is used to generate the 0.1 ... 0.9 bins, would that be most helpful?

-San

[Chris White]

Hi San,

Thank you for your reply and offer to expand the details regarding your code and calculations.

Rather than consume your time further I have spent a lot of time parsing the voila tsv to a new file and interrogating the output against the voila tsv and web browser.

I feel confident now in the data and I think we can close this request off and not take up any more of your time. I assume you are rather busy. 

Thanks again and I will be in touch soon regarding another interesting query? LOL sorry. 

Work well

Regards 

Chris

--
You received this message because you are subscribed to a topic in the Google Groups "majiq_voila" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/majiq_voila/UeRiPrRqdiA/unsubscribe.
To unsubscribe from this group and all its topics, send an email to majiq_voila...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/majiq_voila/d47c2a3f-1931-4a3a-9b2b-0a4910b1f465n%40googlegroups.com.

[San Jewell]

All Good Chris,

Sorry, there has been some firefighting going on lately, but am slowly catching back up :) I'll look out for your other question.

[Chris White]

Hi San,

Hopefully you are okay putting out those fires!

Thanks, I would appreciate you taking a look into my queries. For quick reference here is a link:

https://groups.google.com/g/majiq_voila/c/n3ufFzkd3p0/m/8e1QOk-VBgAJ

Thank you.

Regards

Chris

To view this discussion on the web visit https://groups.google.com/d/msgid/majiq_voila/5937e3d9-8c31-493c-8bfa-cb5f5fd7c76dn%40googlegroups.com.