Closed skchronicles closed 5 months ago
I just submitted a PR to fix the issue here. Thank you again for your time. I appreciate it!
Here is a new screenshot of the KO FLARE A>G results at the TMPO locus:
It is calling enriched A>G regions correctly now (addressing the inf problem fixed everything, see PR for fix).
Thank you so much for posting this fix!
Yes thanks for taking the time to look into this and making the PR @skchronicles , looks good and I will merge it in.
Hello there,
I hope you are having a great day! Thank you again for creating and maintaining this awesome pipeline. It has been extremely useful for a project I have involving A-I editing.
I just wanted to point out an issue I noticed while running FLARE. I will submit a PR shortly, but here is a brief description of the problem below.
I first noticed the issue while viewing some of the output files of SALIOR/FLARE within IGV. In the image(s) below, I have zoomed into a random region. In this figure, the top/bottom coverage tracks are colored by allele frequency.
For this specific project, we were interested in A-I editing sites, so we are looking for A>G editing on the forward strand. These sites are represented as green and yellow coverage bars at any given location.
As you can see here, SAILOR is doing an awesome job calling the A>G sites, and FLARE is also doing an amazing job calling A-I editing regions for the WT samples. But for my KO samples, FLARE did not call this region. After taking a closer look, I noticed something strange in my KO samples.
I tried to grep for this gene within the KO samples' FLARE results, and there were no hits (as expected by the IGV screenshot), so I started looking back further into FLARE's previous output files, and I noticed this had gene
inf
background_rates.So, then I looked into the step that created this output file, and I found where the error was occurring.
In this script on line 45, there is an edge-case where division by 0 can occur:![image](https://github.com/YeoLab/FLARE/assets/18038345/ff63eba6-8f70-40af-8b59-813b377e7ac3)
For my dataset, it happened right here:
For this particular region, the math works out to:
(2-1)/(22-22)
. This causes more issues later on in the script here when finding the mean (resulting in inf values) and then the background_rate calculation (resulting in inf values again).I will submit a PR to fix the issue now. Please feel free to review it whenever you have some free time. Thank you again for your time, and thank you again for creating this awesome pipeline! Please let me know if you have any questions, and have a wonderful weekend!
Best Regards, @skchronicles