nx_vcf.nx

nx_vcf.nx -- VCF (Variant Call Format) single-line writer. license_tier: INDEPENDENT_REDERIVE genealogy_id: international-research-sources/danecek-2011-vcf-spec-v4.2 G2.4 of NISHI_GENOMICS_SUBSTRATE_ROADMAP.md. Closes the variant output side: called variants from snv_call become VCF lines that every downstream tool (bcftools / vcftools / IGV / ClinVar / dbSNP / gnomAD / nf-core pipelines) consumes. VCF v4.2 data-line format (Danecek 2011, NCBI hosts canonical): CHROM POS ID REF ALT QUAL FILTER INFO [FORMAT SAMPLE...] tab-separated, terminated by LF (or CRLF on Windows) G2.4 emits the 8 required fields: CHROM : caller-supplied reference sequence name (no validation) POS : 1-indexed (VCF spec); caller converts from substrate's 0-indexed internal coords ID : "." (no rsID lookup yet; G2.4b composes dbSNP table) REF : single ACGT letter from ref_code (NX_DNA_A/C/G/T) ALT : single ACGT letter from alt_code QUAL : integer Phred-scaled quality (typically 0..60 mapq-equivalent) FILTER : "PASS" if pass_flag=1, "." if 0 INFO : "DP=<depth>" (caller can extend; G2.4 just emits DP) Why a single-line writer (not a streaming serialiser): VCF lines are line-by-line composable. Caller drives the per- variant loop, calls this per variant, accumulates bytes into a buffer or writes them to a file. Keeps this primitive small + composable + testable without I/O surface. What G2.4 does NOT do (deferred): - VCF header (##fileformat=VCFv4.2 + ##INFO + ##FILTER + #CHROM) -- G2.4a - Multi-allelic ALT (A>G,T) -- G2.4b - FORMAT + per-sample fields (GT:AD:DP) -- G2.4c - INFO extension (AF, AC, AN, MQ, MQRankSum, etc.) -- G2.4d - VCF parsing (read existing VCF files) -- G2.4e - bgzip + tabix indexing -- G2.4f - VCF -> SAM rsID annotation lookup -- G2.4g