-
Notifications
You must be signed in to change notification settings - Fork 73
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Precise variants with homology longer than the variant are incorrectly filtered #243
Comments
I'll put in a fix for this in the next GRIDSS release |
The work-around for the current version is to supply a CONFIGURATION_FILE parameter with:
Unfortuantely, that output file will be very large. The variants in question are incorrectly filtered as |
Thank you very much for the response! I will try the configurations you suggested, and see how it works. Look forward the new release! Best, |
Hi Daniel, Thank you for the help! Best, |
The next check I'd do is to gave a look at what GRIDSS assembled. You can
find the contigs in the assembly bam file specified in the `ASSEMBLY`
parameter. If there are assemblies, then have a look at the
assembly.bam.gridss.working/assemblyy.bam.sv.bam file as well. This .sv.bam
contains the assembly contigs after realignment to identify the breakpoint
(ie, the assembly contig 'reads' get converted from soft-clipped reads to
split read).
If the .sv.bam contains split read alignments spanning the event then the
most likely cause is the filtering workaround didn't work.
If the assembly.bam.sv.bam contig alignments don't overlap with the
assembly.bam contig alignments (e.g. they align to a different homologous
region) then this site won't have GRIDSS assembly support.
If there are no assemblies in the assembly.bam file then it's likely that
the read mapping quality scores are too low and GRIDSS is ignoring the
reads. By default this is 20 but can be lowered in the config file.
```
minMapq=10
```
…On Wed, Sep 11, 2019 at 2:01 AM YuanwenGuo ***@***.***> wrote:
Hi Daniel,
I rerun GRIDSS with the configuration settings as you suggested. It helped
a lot and some break-ends can be detected now. But unfortunately, this
bread-end (POS:837244) in the picture could not be detected. I wonder if
there is any other parameters I can adjust to make it work?
Thank you for the help!
Best,
Yuanwen
—
You are receiving this because you were assigned.
Reply to this email directly, view it on GitHub
<#243?email_source=notifications&email_token=ABOBYOFGHIGPYTZ3ZE7ZB43QI7AEZA5CNFSM4IU6SKZKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD6LT35I#issuecomment-530005493>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/ABOBYOFSVSOJGONNYPF27RLQI7AEZANCNFSM4IU6SKZA>
.
|
Hi Daniel, I have tried configurations as following, but they didn't work. #################### Variant calling#################### I wonder if there is any other filtering I can play with? Thank you! Best, |
That is correct.
GRIDSS will report variants even when there is not assembly support (FILTER=NO_ASSEMBLY). That said, GRIDSS heavily favours variants that do have assembly support.
This parameter changes the window size for S/W imprecise homology detection. It's not going to change sensitivity but pushing it out to 2kbp will cause false negatives in the homology alignment due to the lack of flanking anchors. I'd leave this one at the default value.
Have you tried my Can you post the SAM records for assembly.sv.bam record in question, as well as few of the supporting reads for the call that isn't detected? I'm getting quite curious as to why that particular site is so refractory. What's your upstream pipeline look like? Which aligner and additional steps are you performing (e.g markdup BQSR)? |
The hom > variant length should be resolved in 2.6.2. |
Hi Daniel, I tried to attach the original bam file, but seems like bam type is not supported here. Please let me know if the bam file will be helpful for this issue. Yes, I did try minMapq=10, unfortunately this breakend still can not be detected. Thank you very much for the help! Best, |
The upstream pipeline is performed by my colleague. We used bwa and then picard for removing duplicates. Thank you, |
It looks like you have a lot of duplicated sequences in your scaffolds. GRIDSS takes the conservative approach of only calling variants from uniquely mappable sequence. Single breakends are called when one side is multi-mapping but no calls are made when neither side has good mapping quality support. Looking at your reads, they're all either mapq=60, or mapq=0. Furthermore, the read alignments of the mates are split across at least 4 different locations (scaffold_15:2510710, scaffold_15:2739465, scaffold_49:387751, scaffold_51:530826). GRIDSS doesn't make a call since whichever position it guesses is likely to be wrong. If you really want the 4 calls made (e.g. you're looking for misassemblies in your scaffolds), then you can set the minMapq=0, and deal with the much higher false discovery rate. You'll also need to add |
Thank you for the detailed explanation! Now I can see why this SV is not detected. GRIDSS is a really powerful breakend detector from our experience, much better than some other popular programs (thanks a lot for developing such a great tool). At this point, we probably don't want to increase false positives, and will only take what GRIDSS reports which are reliable calls. Best, |
Hello,
![example](https://user-images.githubusercontent.com/26492070/64539784-b86c6a00-d2e4-11e9-86af-325bfc5808b5.png)
I am trying to use GRIDSS detecting possible breakends in multiple samples. It successfully identifies most of breakends, but looks like some breaks can not be picked up with default configuration settings. Take one breakend as example:
I tried to adjust configuration with the following parameters, but it didn't help much:
minMapq = 10
assembly.minReads = 1
variantcalling.minReads = 1
variantcalling.minScore = 10
variantcalling.minSize = 10
I also tried to use cohort scripts to process my samples, which did not detect the breakends too.
Could you please give me any suggestions about how to change configurations to make GRDISS be able to detect these kinds of breakends?
Thank you!
Best,
Yuanwen
The text was updated successfully, but these errors were encountered: