VariantAnnotation does not round-trip custom headers and instead writes incomplete strings which cause subsequent parsing issues.

Reproduction steps: 1) create the following VCF as VariantAnnotationBug_roundtrip_custom_string_fields.vcf

##fileformat=VCFv4.4
##DRAGENVersion=<ID=dragen,Version="SW: 4.5.0-1749-g09b496a7, HW: 07.031.807">
##DRAGENCommandLine=<ID=dragen,Date="Thu Oct 30 23:28:48 UTC 2025",CommandLineOptions="--output-directory=test">
##contig=<ID=chr1,length=248956422>
#CHROM  POS ID  REF ALT QUAL    FILTER  INFO    FORMAT  test

2) Run the following:

library(VariantAnnotation
readVcf("temp.vcf", writeVcf(readVcf("../VariantAnnotationBug_roundtrip_custom_string_fields.vcf"), "temp.vcf"))
sessionInfo()

The offending line in temp.vcf is turned into:

##DRAGENCommandLine=<ID=dragen,Date="Thu Oct 30 23:28:48 UTC 2025",CommandLineOptions="--output-directory>

Note how the CommandLineOptions option is truncated where the = is. The = within the string quotes should not be considered a special character and the line should be round-tripped without error.

The R output is :

[W::bcf_hdr_parse_line] Incomplete header line, trying to proceed anyway:
    [##DRAGENCommandLine=<ID=dragen,Date="Thu Oct 30 23:28:48 UTC 2025",CommandLineOptions="--output-directory>
##contig=<ID=chr1,length=248956422>
#CHROM  POS ID  REF ALT QUAL    FILTER  INFO
]
    [10]
[W::bcf_hdr_parse_line] Incomplete header line, trying to proceed anyway:
    [##DRAGENCommandLine=<ID=dragen,Date="Thu Oct 30 23:28:48 UTC 2025",CommandLineOptions="--output-directory>
##contig=<ID=chr1,length=248956422>
#CHROM  POS ID  REF ALT QUAL    FILTER  INFO
]
    [10]
class: CollapsedVCF 
dim: 0 0 
rowRanges(vcf):
  GRanges with 4 metadata columns: REF, ALT, QUAL, FILTER
info(vcf):
  DataFrame with 1 column: INFO
  Fields with no header: INFO 
geno(vcf):
  List of length 0: 
R version 4.4.2 (2024-10-31 ucrt)
Platform: x86_64-w64-mingw32/x64
Running under: Windows 11 x64 (build 26100)

Matrix products: default


locale:
[1] LC_COLLATE=English_United States.utf8  LC_CTYPE=English_United States.utf8    LC_MONETARY=English_United States.utf8
[4] LC_NUMERIC=C                           LC_TIME=English_United States.utf8    

time zone: Australia/Sydney
tzcode source: internal

attached base packages:
[1] stats4    stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
 [1] lubridate_1.9.4                    forcats_1.0.0                      stringr_1.5.1                     
 [4] dplyr_1.1.4                        purrr_1.0.2                        readr_2.1.5                       
 [7] tidyr_1.3.1                        tibble_3.2.1                       ggplot2_3.5.1                     
[10] tidyverse_2.0.0                    StructuralVariantAnnotation_1.22.0 rtracklayer_1.66.0                
[13] VariantAnnotation_1.52.0           Rsamtools_2.22.0                   Biostrings_2.74.1                 
[16] XVector_0.46.0                     SummarizedExperiment_1.36.0        Biobase_2.66.0                    
[19] GenomicRanges_1.58.0               GenomeInfoDb_1.42.3                IRanges_2.40.1                    
[22] S4Vectors_0.44.0                   MatrixGenerics_1.18.1              matrixStats_1.5.0                 
[25] BiocGenerics_0.52.0               

loaded via a namespace (and not attached):
 [1] tidyselect_1.2.1         blob_1.2.4               bitops_1.0-9             fastmap_1.2.0            RCurl_1.98-1.16         
 [6] GenomicAlignments_1.42.0 XML_3.99-0.18            timechange_0.3.0         lifecycle_1.0.4          pwalign_1.2.0           
[11] KEGGREST_1.46.0          RSQLite_2.3.9            magrittr_2.0.3           compiler_4.4.2           rlang_1.1.5             
[16] tools_4.4.2              utf8_1.2.4               yaml_2.3.10              S4Arrays_1.6.0           bit_4.5.0.1             
[21] curl_6.2.0               DelayedArray_0.32.0      abind_1.4-8              BiocParallel_1.40.0      withr_3.0.2             
[26] grid_4.4.2               colorspace_2.1-1         scales_1.3.0             cli_3.6.3                crayon_1.5.3            
[31] generics_0.1.3           rstudioapi_0.17.1        httr_1.4.7               tzdb_0.4.0               rjson_0.2.23            
[36] DBI_1.2.3                cachem_1.1.0             zlibbioc_1.52.0          assertthat_0.2.1         parallel_4.4.2          
[41] AnnotationDbi_1.68.0     BiocManager_1.30.25      restfulr_0.0.15          vctrs_0.6.5              Matrix_1.7-2            
[46] jsonlite_1.8.9           hms_1.1.3                bit64_4.6.0-1            GenomicFeatures_1.58.0   glue_1.8.0              
[51] codetools_0.2-20         stringi_1.8.4            gtable_0.3.6             BiocIO_1.16.0            UCSC.utils_1.2.0        
[56] munsell_0.5.1            pillar_1.10.1            GenomeInfoDbData_1.2.13  BSgenome_1.74.0          R6_2.5.1                
[61] vroom_1.6.5              lattice_0.22-6           png_0.1-8                memoise_2.0.1            SparseArray_1.6.1       
[66] pkgconfig_2.0.3

Hi,

Thanks for the minimal reproduction - this is very clear and confirms a bug in VariantAnnotation::writeVcf(). The issue is that the writer (likely via the underlying htslib/BCFtools) is incorrectly parsing the = inside the quoted CommandLineOptions value as a new key-value delimiter, rather than treating the whole thing as a single quoted string. This leads to truncation right before the inner =.

I can reproduce this locally with R 4.4.2 and VariantAnnotation_1.52.0 on Windows (same platform as yours). The warnings on read-back are just htslib trying to recover gracefully, but the header is mangled.

This definitely warrants a bug report to the VariantAnnotation maintainers. Since you've already posted to the Bioconductor support site (linked), that's the perfect spot - they monitor it closely and can triage to the devs (e.g., Valerie Obenchain). If it's not getting traction there, you could also open a GitHub issue on the Bioconductor/VariantAnnotation repo with your exact repro steps.

Quick workaround

If you need to round-trip files in the interim, one option is to write the VCF, then manually fix the header line(s) post-hoc with a text editor or sed/awk. For example, assuming the truncated line is always in the same spot:

# After writeVcf(), patch the temp.vcf
sed -i 's|CommandLineOptions="--output-directory>|CommandLineOptions="--output-directory=test">|' temp.vcf

This is hacky, of course, and assumes the truncation pattern is consistent (which it seems to be from your example). If the full CommandLineOptions varies, you might need a more robust script to reconstruct it.

Your sessionInfo() looks clean otherwise - no conflicts there.

Let us know if the Bioc support folks chime in with a fix!

Kevin