lofreq call ran into buffer overflow

[NCC-2310-Anna]

Hi there,

I'm working with lofreq to call variants in a simulated bacterial dataset, and in some cases, lofreq call runs into a buffer overflow.
That's the command I used for calling:
lofreq call -f "path/to/reference/fasta" --call-indels -o "path/to/output.vcf" "path/to/bam/file.bam" 2> "path/to/log/file.log"

And that's the error message I've got:
*** buffer overflow detected ***: terminated

I've already inspected the BAM files, but they look normal. I'm using lofreq version 2.1.5, installed via conda, on Ubuntu 24.04.2 LTS. Has anyone had similar issues or/and maybe a workaround?

Thank you for your time and assistance!

[andreas-wilm]

Hi there, that should never happen and is a programming bug. Is there any chance you could provide example input files?

[NCC-2310-Anna]

Hi,
Thanks for the fast response.
My dataset was too big for GitHub Upload. I've uploaded it to my university cloud: Link.
filtered.bam is the raw file, I've processed with the indelqual command first. That ran without any errors.
The original reads in the datasets where lofreq crashed were always 300 bp, paired-end.

[andreas-wilm]

Thanks! I can reproduce this and it happens exactly at position 965384 and nowhere else. A minimal (non-)working example is: lofreq call -f reference.fasta -o output.vcf indelqual.bam -r NZ_CP030271.1:965384-965384. At first glance there is nothing wrong with this position. I will have to look at it in detail.

For now, can I suggest the following hack:

# run on first part of NZ_CP030271.1
lofreq call  -f ref.fa --call-indels -r NZ_CP030271.1:1-965383 -o out1.vcf  file.bam
# run on second part of NZ_CP030271.1
lofreq call  -f ref.fa --call-indels -r NZ_CP030271.1:965385-4000000 -o out2.vcf  file.bam
# run on all of NZ_CP030272.1
lofreq call  -f ref.fa --call-indels -r NZ_CP030272.1:1-4000000 -o out3.vcf  file.bam

The concatentate the three vcf files (bcftools concat I think).

This will give you near identical result to running on the entire BAM file, but skips the problematic region.

[andreas-wilm]

Further debugging info:

Looks like the source is a strcpy in mpileup.

The problem doesn't appear on filtered.bam

Removing compiler optimization makes the bug disappear.

[NCC-2310-Anna]

Thanks for the debugging info!
Is there any possibility that you could make a debugged version available via BioConda?

[andreas-wilm]

Yes, once I find the bug :)