juncBASE (Junction-Based Analysis of Splicing Events) bugs?
1
1
Entering edit mode
6.9 years ago
Anna S ▴ 500

Hello,

People I work for love the juncBASE output, but I've been having major issues with the software:

1. I have only gotten it to work to display the correct genes when I use UCSC mapping, etc. (which has 'chr' prefix) but not  with Ensembl.  When I use Ensembl most events are considered N (novel) rather than K (known), and some genes listed are clearly wrong with wrong chr, etc.

2. There are thread/process issues so that some of the samples never complete.  Worst, the last samples listed sometimes are shown to have completed but have different results than if they were run somewhere in the middle of the command.  I suspect this problem is due to thread/process issues with the last 2 samples.

3. I could deal with 2. above by running it, saving the results and then swapping the last samples with some in the middle of the command and rerunning to get the correct results for the last 2 samples, but sometimes even samples listed in the middle have different results each time juncBASE is run.

I have tried to contact the author, but usually these conversations take months of back and forth and nothing ever gets resolved.  I was wondering if anyone else has run into these issues and has any insight into the matter?

I tried to use MATS instead but my customers didn't like the output.

I'm seriously considering writing a new software as this one seems unusable?

juncBASE Alternative Splicing RNA-seq • 1.9k views
0
Entering edit mode
6.6 years ago

Hi there,

I have the same no 1 problem like you and I don't know what to do with it, I am working on arabidopsis and ensemble genome from TAIR10 is the only one available with no UCSC annotation to work on. So all events are always annotated as Novel N.

For problem 2 and 3, I believe that you just to keep use --check as the author recommended. Sometimes may be the ChrPt or so is absent because there is no reads coverage in those files

I have contacted the author but haven't heard from her so far.

Hope if you get anything to recommend it to me.

Thanks

0
Entering edit mode

I experience often the problem that you mentioned where due to thread/process problems juncBASE exits but some of the samples are not fully processed (based on --check) and one needs to keep rerunning juncBASE several times until all of the samples are completely finished (if they're not finished one can also determine this (besides --check) by noting that the temporary files were not deleted). What I mentioned in 2-3 above is more serious, in that I have run into a problem where --check claims that the sample is finished, and in fact, all of the temporary files were deleted, but the resulting output is incomplete compared to a previous identical run. I have not run into this problem often, but the day I wrote this post I ran into it several times and that is why I was frustrated enough to write it!!