issue in Create PanGenie-ready VCF from Minigraph-Cactus VCF

[ChuanzhengWei]

Hi, eblerjana

I used Minigraph-Cactus to construct a graph pangenome. Following the instructions from this page: prepare-vcf-MC, I performed some preprocessing steps using the provided scripts. The commands I ran are as follows:

vg convert -f xx.gbz -t 12 > xx.gfa
vg deconstruct -P HYZ -H '#' -e -C -a -t 24 xx.gfa|bgzip -@ 48 > xx.c.vcf.gz
vcfbub -l 0 -r 100000 --input xx.c.vcf.gz | bgzip > xx.vcfbub.vcf.gz

bcftools +fill-tags -Oz xx.vcfbub.vcf.gz -o xx.f.vcf.gz --threads 12 -- -t AN,AC,AF 
bcftools view xx.f.vcf.gz --min-ac 1 | python3 filter-vcf.py 0.8 2> filtered.log 1> filtered.vcf
python3 annotate_vcf.py -vcf filtered.vcf -gfa xx.gfa.gz -o filtered_ids-tmp &> annotate.log

However, when running annotate_vcf.py, I encountered the following error:

Preprocessing...
Reading sequence information from GFA file...
Done reading GFA.
Annotate the VCF file...
Traceback (most recent call last):
  File "annotate_vcf.py", line 453, in <module>
    multi_line, bi_lines = decompose(line, gfa)
  File "annotate_vcf.py", line 337, in decompose
    ref_pos = get_ref_position(allele[0], gfa, add_flank)
  File "annotate_vcf.py", line 277, in get_ref_position
    position = int(gfa[node.id][0]) + len(gfa[node.id][1])
TypeError: int() argument must be a string, a bytes-like object or a real number, not 'NoneType'

Interestingly, if I first filter the VCF file using bcftools view -m2 -M2 before passing it to annotate_vcf.py, the script runs without any issues.

Additionally, when I use a VCF file converted from a Minigraph-Cactus GBZ file, modifying the genotype format before using it as input for PanGenie-index, I encounter another error:

Determine allele sequences ...
Read reference genome ...
Found 10 chromosome(s) from the reference file.
Read input VCF ...
GraphBuilder: skip variant at Chr01:12 since variant is less than 2 * kmer size from start or end of chromosome. 
terminate called after throwing an instance of 'std::runtime_error'
  what():  GraphBuilder: variant at Chr01:939872 overlaps previous one. VCF does not represent a pangenome graph.

Aborted

Could you please help me troubleshoot these issues? Any guidance or suggestions would be greatly appreciated.

Thank you for your time and support!

Best regards

[eblerjana]

Hi, the error:

 Preprocessing...
Reading sequence information from GFA file...
Done reading GFA.
Annotate the VCF file...
Traceback (most recent call last):
  File "annotate_vcf.py", line 453, in <module>
    multi_line, bi_lines = decompose(line, gfa)
  File "annotate_vcf.py", line 337, in decompose
    ref_pos = get_ref_position(allele[0], gfa, add_flank)
  File "annotate_vcf.py", line 277, in get_ref_position
    position = int(gfa[node.id][0]) + len(gfa[node.id][1])
TypeError: int() argument must be a string, a bytes-like object or a real number, not 'NoneType'

indicates that there was no SO:i tag available for a node in the graph that according to the VCF, should be a node on the reference genome backbone. Are you sure that the VCF is relative to the same reference that was used as a backbone for the MC graph?

I'm not sure I understand how the VCF looks like that you mention in the second question, but I assume that the VCF you are providing to PanGenie was not yet preprocessed with vcfbub to remove non-top level bubbles? This is necessary before applying PanGenie.

[ChuanzhengWei]

Hi, eblerjana

Thank you so much for your response! After processing the VCF generated by the MC pipeline with vcfbub, everything seems to be working fine now.

Do you think it is still necessary for me to troubleshoot the issue with annotate_vcf.py? I checked and found that the VCF file was generated using the same reference genome as the backbone for the MC graph. However, the error persists, and the generated VCF file stops at the first non-biallelic site. This is the final VCF file I obtained.

filtered_ids-tmp.vcf.gz
filtered_ids-tmp_biallelic.vcf.gz

Looking forward to your advice!

[eblerjana]

It depends on what you want to do. For genotyping with PanGenie itself, the annotations are not needed. They are only necessary if one wants to convert the bubble genotypes generated by PanGenie to genotypes of all variants nested inside of the bubbles (see Wiki https://github.com/eblerjana/pangenie/wiki/A:--Genotyping-variation-nested-inside-of-bubbles).

It still seems to me that the problem is missing SO:i annotations in the graph. All nodes in the graph that are traversed by the reference backbone must have these annotations, however the error indicates that these are missing. Can you double check that all reference nodes indeed have these annotations in the GFA file?

[ChuanzhengWei]

Thank you for your suggestion. I will follow your advice and provide timely feedback.

[Han-Cao]

Hi,

I also had this issue, and I found it is because you are generating a GFA not an rGFA. The SO:i annotation is only included in rGFA.

vg convert -f xx.gbz -t 12 > xx.gfa

I think you can try to use -Q to specify the reference genome and generate an rGFA.

See help in vg convert:

    -Q, --rgfa-prefix STR  write paths with given prefix as rGFA tags instead of lines
                           (multiple allowed, only rank-0 supported)