Error running vg construct on GIAB_0.6 sample ->> error[VPKG::load_one]: Correct input type not found while loading handlegraph::MutablePathDeletableHandleGraph

[rbr7]

Hi,

I am trying to run and replicate the giraffe SV paper results for GIAB 0.6 sample, to genotype SVs from the Illumina paired reads. I am currently following the steps from the giraffe paper, where instead of using the toil-vg scripts, the steps mentioned in the Snakefile for giraffe paper are the ones i am referring to (https://github.com/vgteam/vg_snakemake/blob/master/Snakefile).

Input ref : hs37d5.fa.gz and its corresponding index
input vcf : HG002_SVs_Tier1_v0.6.vcf.gz and its index file
After running the "vg construct" step without pre-processing input VCF, I am getting the following error in the log files,

"stderr": "Restricting to 1 from 1 to end\n building graph for 1 [ ] 0.0%\rwarning:[vg::Constructor] Lowercase characters found in 1; coercing to uppercase.\nwarning:[vg::Constructor] Unsupported IUPAC ambiguity codes found in 1; coercing to N.\nerror:[vg::Constructor] unacceptable characters found in 1.\nerror[VPKG::load_one]: Correct input type not found in standard input while loading handlegraph::MutablePathMutableHandleGraph\n",

->> I figured out the initial warnings, but the error towards the end re occurs, even if I re-Index the the input files.
->> Hickey et al 2019, uses the GIAB 0.5 samples as here, (https://github.com/vgteam/sv-genotyping-paper/tree/master/human/giab),

but the pre-processing step for preparing the input SV catalog file, is what gives me null VCF with only headers when running for GIAB 0.6, similar to steps below:

wget -nc ftp://ftp-trace.ncbi.nlm.nih.gov/giab/ftp/data/AshkenazimTrio/analysis/NIST_UnionSVs_12122017/svanalyzer_union_171212_v0.5.0_annotated.vcf.gz
wget ftp://ftp-trace.ncbi.nlm.nih.gov/giab/ftp/data/AshkenazimTrio/analysis/NIST_UnionSVs_12122017/svanalyzer_union_171212_v0.5.0_annotated.vcf.gz.tbi
vcfkeepinfo svanalyzer_union_171212_v0.5.0_annotated.vcf.gz NA | vcffixup - | bgzip > giab-0.5.vcf.gz
tabix -f -p vcf giab-0.5.vcf.gz

->> What I am trying to understand is how to prepare the input SV catalog file for GIAB 0.6 to be used in the "vg construct - ..." step, so as to avoid the above errors ?

[adamnovak]

I think your problem is:

[vg::Constructor] unacceptable characters found in 1.

That then causes the second complaint from the second command in the pipeline: the first command died without outputting a graph, and the second command died because it consequently can't read the graph that the first one didn't make.

I think your reference file has characters other than ACGTN or the supported ambiguity codes MRWSYKVHDB. Do you have maybe - or * or something else weird in your input FASTA sequences somehow?

I don't think your problem has anything to do with the VCFs. In general terms, you want to prepare the VCFs so that they as much as possible specify all the variants with base-level precision, and so that any symbolic descriptions of are fully self-consistent (i.e. starts, lengths, ends, and sequences agree). You also want to make sure that overlapping VCF records don't conflict with each other about certain samples (like reporting a heterozygous deletion overlapping with a heterozygous SNP). Ideally, overlapping variants would be phased or pre-merged.