Open Coppini opened 3 years ago
Running with artic-tools==0.3.1
fixes the artic-tools check_vcf
error on artic==1.3.0-dev
. It might be just a matter of updating the artic-tools
dependency to the newest version.
On v1.2.1, the problem is on Longshot
On v1.3.0, this is actually a duplicate of https://github.com/artic-network/fieldbioinformatics/issues/88
The above error is occurring regularly using ARTIC v1.2.4 in samples with very low coverage (roughly 0 to 5 aligned reads).
Command failed:longshot -P 0 -F -A --no_haps --bam sample_name.primertrimmed.rg.sorted.bam --ref primer-schemes/SARS-CoV-2/V5.3.2/SARS-CoV-2.reference.fasta --out sample_name.merged.vcf --potential_variants sample_name.merged.vcf.gz
As mentioned above, a solution would be to remove the '-A' (automatic max coverage) flag from the longshot command or to make it optional.
When running Artic on a sample with too few reads, resulting in too small coverage, Longshot automatically calculates the maximum coverage as 0, and fails. We're having issues with that in an automatic pipeline, where some of our samples keep failing due to that, since they have too few reads (as expected, since they were negative samples).
Already reported it to them at https://github.com/pjedge/longshot/issues/72, but I agree with them that the best solution would be to simply not use the
-A
flag (or, alternatively, to ignore the error in the Artic pipeline when it happens, and create an empty file instead).I tried running the branch in
1.3.0-dev
as well, but end up with an error in theartic-tools check_vcf
step:It seems the problem appears when the VCF file is empty:
Running
artic-tools check_vcf
ends up on aSegmentation fault (core dumped)
when reading the empty VCF: