Discrepancies between voila tsv and voila view after heterogen analysis and problems with voila tsv output

[Miriam Martínez]

Hi all,

While checking the results provided by voila tsv and voila view, I have seen that there are discrepancies between both tools. I will use the first gene that appears on my result list (independently if it is significant/changing) that is LINC00824. On the file voila_tsv_result.png is a screenshot of the result from voila tsv. 

First of all, the problem I've seen with voila tsv output is, that if youtake a look on the BD_median_psi and the CTLmedian_psi columns, it can be seen that only 2 values are shown as result, while if we look at infoscore and mannwhitneyu columns, we can see 4 results. Is this possible? Shouldn't there be 4 values on BD_median_psi and CTLmedian_psi columns too? This problem happens with other genes on my list as well.

Secondly, related to the discrepancy between voila tsv and voila view, if you look at voila_view1.png file (which is a screenshot of part of the LSV related to LINC00824), you can see that that the mannwhitneyu value that appears for that subevent doesn't coincide with any of the values that appear on the voila tsv output (this happens with almost every subevent of that anf other LSVs and with other stat tests). 

Also, I was wondering if there is a way to obtain the deltapsi values that appear on voila view, as the deltapsi that appears there seems to be the minus of the E(PSI) of each group and those values don't appear on the tsv file (or at least I haven't seen them). I have tried to minus the values on the BD and CTL mdeian_psi columns to try to get a dPSI value for my data too, but as it differs from the one that appears on voila view, I'm not sure anymore how I should calculate the dPSI values for my results.

Note: my data is corrected with the moccasin step in case this info is useful.

Thank you very much for your help.

Best,

Miriam

[San Jewell]

Hi Miriam, 

I would again like to break this into components so that I can make sure I address everything, but we are again in a case where there were some bugs to fix. 

1) Differing count of stat vs median psi columns: For both the median psi and the percentile psi columns, only the first two junctions values were shown instead of all of them. This has now been fixed. 

2) Stat values differ between tsv file and voila view. The issue was a duplicate column name, which ended up overwriting the stat value with the stat_quantile value, causing the difference and for the stat_quantile column to be empty. This has now been fixed

3) Currently in het mode, a simple subtraction is the way that dpsi values are obtained. In my testing I was able to get the same value as dpsi by subtracting median psi values (which can also be viewed in view mode by hovering over the group name under the violin plot) -- I believe you will likely see it is correct but you were likely tripped up by the bug where only the first two junctions were shown in the tsv. However, I am not completely sure of this, and I was wondering if you could specifically show a case where the dpsi is not the same as the subtraction. After this, I don't think it would be a problem to add a column to the tsv output for the subtraction result. Let me know. 

Thanks, 

-San

[Miriam Martínez]

Hi San,

Finally I was able to try and take a look at the tsv output after the updates:

1) Now appear more than 2 values in  both the median psi and the percentile psi columns when it should. One thing I would like to point out too is that, now in the tsv the genes that have no data appear also in the tsv but the values are filled with "nan" (without the --show-all flag). I attach a screenshot as an example (example_nans_voila.png). It isn't a severe issue as the info can be easily filtered, but I think is worth mentioning it.

2) I have checked some examples and now the values from the tsv and voila view coincide.

3) Now the  dpsi values coincide with the substraction of the median_psi values.

As you said, probably that bug that you have corrected made all these discrepancies, but now all seems to work as it should, so good news! Also, it would be great if you could add that extra column with the substraction result. Thanks for all!

Best,

Miriam

[San Jewell]

Hi Miriam, 

I'm glad it was resolved. The dpsi subtraction column will be in the next minor update. 

Thanks!

-San