PBSV_CALL inserts NUL bytes (^@) in the middle of a record?

[MikeHala]

Description of the bug

Hi there,
I am running pacvar version 1.0.1 on HPC and tried to process PacBio data for HG002 (downloaded from https://downloads.pacbcloud.com/public/2026Q1/HG002-SPRQ-Nx/EA260902-Use1/).

My command line is like this

./nextflow run ..../pacbio/nf-core-pacvar/1_0_1/main.nf \
--input ./samplesheet.csv --outdir ./results \
--genome 'GATK.GRCh38' -profile eddie \
--workflow wgs --snv_caller deepvariant \
--skip_demultiplexing \
-c ./custom.config \
-c ./base.config \
-resume

Alignment completed successfully, still running SNV calling, but got a problem with the SV calling.
PBSV_DISCOVER and PBSV_CALL finished successfully, but NFCORE_PACVAR:PACVAR:BAM_SV_VARIANT_CALLING:BCFTOOLS_INDEX crashed with

Command error:
  INFO:    Converting SIF file to temporary sandbox...
  [E::hts_idx_push] Invalid record on sequence #72: end 1 < begin 138271
  index: failed to create index for "HG002.sv.vcf.gz"
  INFO:    Cleaning up image...

Upon closer examination of the HG002.sv.vcf file there appears to be an extra field introduced by PBSV_CALL for the middle record, just before the REF field

awk 'NR>=56885 && NR<=56887 {print}' HG002.sv.vcf
chrUn_KI270519v1        137207  pbsv.INS.53107  T       TGGAATAGAATGGAGTGGAGTGGAATGGAATGGAGTGGAGTGGAGTGGAATGGAGTGGAGTGGAGTGGAATGGAATGGAGAGGAGTGCAGA     .       NearContigEnd   SVTYPE=INS;END=137207;SVLEN=90  GT:AD:DP        1/1:31,200:231
chrUn_KI270519v1        138271  pbsv.INS.53108          TGGAGTGGAACGCAGAGAATGGAATGGAG   .       NearContigEnd   IMPRECISE;SVTYPE=INS;END=138271;SVLEN=29        GT:AD:DP        0/1:45,80:125
chrUn_KI270515v1        2949    pbsv.DEL.53109  ATTGTAGAAAAGGAAATGTCATCAAATAAATACTACACAGAAGCATTCAGAAAAACTTCTTTGTTATGAGTGCATTCATCACACAGAGTTGAACCTTTCCTTTGACTGAACAGTTTTGAAACACTCTTTTTGCAGAATCTGCAAGTGCATATTTTAGAGCTTTGAGGACAA     A       .       PASS    SVTYPE=DEL;END=3119;SVLEN=-170  GT:AD:DP      0/1:95,26:121

Shown in another way

cat HG002.sv.vcf | sed -n '56885,56887p' | cat -A
chrUn_KI270519v1^I137207^Ipbsv.INS.53107^IT^ITGGAATAGAATGGAGTGGAGTGGAATGGAATGGAGTGGAGTGGAGTGGAATGGAGTGGAGTGGAGTGGAATGGAATGGAGAGGAGTGCAGA^I.^INearContigEnd^ISVTYPE=INS;END=137207;SVLEN=90^IGT:AD:DP^I1/1:31,200:231$
chrUn_KI270519v1^I138271^Ipbsv.INS.53108^I^@^I^@TGGAGTGGAACGCAGAGAATGGAATGGAG^I.^INearContigEnd^IIMPRECISE;SVTYPE=INS;END=138271;SVLEN=29^IGT:AD:DP^I0/1:45,80:125$
chrUn_KI270515v1^I2949^Ipbsv.DEL.53109^IATTGTAGAAAAGGAAATGTCATCAAATAAATACTACACAGAAGCATTCAGAAAAACTTCTTTGTTATGAGTGCATTCATCACACAGAGTTGAACCTTTCCTTTGACTGAACAGTTTTGAAACACTCTTTTTGCAGAATCTGCAAGTGCATATTTTAGAGCTTTGAGGACAA^IA^I.^IPASS^ISVTYPE=DEL;END=3119;SVLEN=-170^IGT:AD:DP^I0/1:95,26:121$

Any advice on how to resolve this would be greatly appreciated!
Alternatively, do you have an ETA for PacVar version 1.1.0?

Best regards,
Mike

Command used and terminal output

Relevant files

No response

System information

No response

[chaochaowong]

@MikeHala Thank you for your report. I encountered the same issue with pbsv: in some alternative contigs, such as chrUn_KI270519v1, the output VCF is malformed. I manually fixed it by replacing "^@" by "." and then run hiphase manually. It was a painful process. Below is what is should look like. I don't know if it is pbsv's bug or there is an space, instead of N, in the fasta file.

chrUn_KI270519v1        138271  pbsv.INS.53108          TGGAGTGGAACGCAGAGAATGGAATGGAG   .
->
chrUn_KI270519v1        138271  pbsv.INS.53108     .    TGGAGTGGAACGCAGAGAATGGAATGGAG   .

The sawfish module has been integrated into v1.1.0dev on the dev branch and can be used with: nextflow run nf-core/pacvar -r dev. BUT I am not sure whether sawfish will work in this particular case, as I have not tested it yet. I will do it next week during the summit.

I am also still adding new features, including annotation support (VEP and AnnotSV) and a Fiber-seq workflow (fibertools/add_nucleosome). I am hoping to make the release after the Nextflow Summit, once I have had a chance to gather feedback and advice from the community.