Voila view plots

[Miriam Martínez]

Dear biociphers team,

I hope everything's going well with you all. As I continue to explore your tool with my data, some new doubts arise and I wanted to ask here seeking for some help. Just to provide some context, I'm using the 'heterogen' option for my analysis.

Firstly, when I check out the heatmaps that appear next to the plots, I've noticed that the statistical value (WILCOXON in my case) matches what's in the corresponding TSV files for each event. However, when I look at the dPSI value, it differs from the substraction of the median of PSI of each group that voila tsv reports. As far as I know, dPSI should be the difference between the two groups, please correct me if I'm wrong, so I was wondering why the values reported on the heatmaps on the violin plots don't correspond to that substraction.

Secondly, I mentioned this in a previous message, but I'm still struggling to figure out how to extract the PSI values that are shown on the violin plots. I apologize if you are already working on this as I asked a time ago, but it would be very helpful to know how to extract this PSI values that MAJIQ calculates.

Thanks a bunch for your help!

Miriam Martínez

[San Jewell]

Hello Miriam,

I did a simple test on majiq het, running three samples per group for two groups. I looked at the median_psi columns for groups 1 and 2 in the TSV, subtracted them, and compared that value to the value when hovering over the upper diagonal of the heatmap. In my first test, the values match. Would you be able to provide an example of the data subset / build config you are using to produce the discrepancy you are seeing? I'll need to figure out how to reproduce it before I can determine if there is some odd edge case where this happens.

Thank you!

-San

[Miriam Martínez]

Hi San, 

sorry for the late response. Just taking a look to the first event reported in the tsv, the values don't match. Example ("voila_version": "2.4.dev102+g2cae150"):
gene_name: GATAD2B
Affected_median_psi: 0.9945;0.0055
Unaffected_median_psi: 0.9292;0.0708

0.9945 - 0.9292 = 0.0653

Meanwhile, voila reports 0.026 (image attached).

I don't know if you need any more information. If so, please let me know.

Also, on the first message there was a second question: " I'm still struggling to figure out how to extract the PSI values that are shown on the violin plots." If you could please help me with this issue too. I need the PSI per sample values to continue the project I'm working on right now, so I would be really gratful if you could guide me with this too.

Thanks!

Miriam

[San Jewell]

Hello Miriam,

(note, google having problems pasting images inline, so I'm putting them as attachments below)

I've tried to reproduce your issue using majiq_academic, which also shows version (t1 image). I looked at a sample let here which shows up as the following (t2 image) , and when running "voila tsv" I get the row for it (t3 image). To me it looks like all values match as expected, so I'm guessing some sort of odd edge case is happening with your run. The easiest way for me to reproduce it would be to start with the same data/config file/list of commands you are using. Do you think it would be possible to provide them to me for debugging purposes?

In regards to the second question, per-sample psi is not currently provided in $ voila tsv or $ voila modulize -- however, I've brought it up with some other lab members today and there is no overt disinterest, so I'd be able to include it in the next release / beta early access version, depending on confirmation by my supervisor. Note that this will likely make the resulting tsv file extremely "wide" -- some software such as microsoft excel may not be able to open all of the columns successfully.

Thanks,

-San

[Miriam Martínez]

Hi San,

after talking to my supervisor, he has given me green light to provide you the files you ask for. Can you specify me what tipe of initial files you need me to provide you apart from the config file and the commands (bams, sj...)? Also I would ask you if you could provide me an email for sending my data, as they are not meant to be public.

Regarding the individual psi per sample, it would be great to have this option, and it wouldn't be a problem to not be able to open the tsv with excel as long as it can be managed with linux.

Thank you very much for your help! I look forward to your response.

[San Jewell]

Hi Miriam,

Thank you for taking the time, I really appreciate you helping me to resolve this bug case and I'm sure it may help others in the future. For the data transfer steps I've sent you an email privately to coordinate. I'll post back here the findings after I'm able to look at the data.

Thanks!

-San