gigwa-import-and-qc

Gigwa import + QC

Overview

Load a genotype dataset into a Gigwa instance and immediately check whether the import is trustworthy. This chains a single import call with the full read-only QC suite and an import-quality audit, then ends with a clean / not-clean judgement. It drives the gigwa-mcp MCP server's tools — the tools do the work; this skill sequences them.

When to use / when not

• Use when the user is ingesting new data: a DArTseq SNP/Silico xlsx report or a .vcf/.vcf.gz, and wants confidence the import worked.
• Don't use to re-QC a run that is already imported — use gigwa-qc-triage instead (same QC steps, no import). For diversity/structure after QC passes, use gigwa-diversity-report.

Prerequisites

• The gigwa-mcp server is connected and gigwa_server_info succeeds.
• Import writes data, so the server must be authenticated: GIGWA_USER / GIGWA_PASS (anonymous access is read-only and cannot import).
• Input files are on a path the server process can read.
• Decide the target coordinates up front: module (database), project, run.

Decide the branch

Input	Tool	Key file param
DArTseq xlsx (SNP and/or SilicoDArT)	import_dartseq	snp_xlsx, silico_xlsx
VCF (.vcf / .vcf.gz)	import_vcf	vcf_path

Optional: genome-anchor DArTseq first

DArTseq tags are unplaced by default. To assign genomic positions, align the tag sequences to a reference before import:

1. map_dartseq_to_reference(snp_xlsx, reference_fasta, min_mapq=20, preset="sr", backend="auto") → writes dartseq_positions.csv.
2. Pass that CSV back as import_dartseq(..., positions_csv="…/dartseq_positions.csv"), or let import_dartseq(..., reference_fasta=…) do the alignment inline.

Skip this whole step for VCF (already positioned) or when unplaced markers are fine.

Step-by-step workflow

1. Import.
   • DArTseq: import_dartseq(module, project, run, snp_xlsx=…, silico_xlsx=…, ploidy=2, skip_monomorphic=False, clear_project_data=False).
   • VCF: import_vcf(vcf_path, module, project, run, ploidy=2).
   • Long imports: pass wait=False to get a progress token, then poll with get_import_progress(progress_token) and cancel with abort_import(progress_token) if needed.
2. Confirm counts with list_variant_sets(). Read back the new run's variant and call-set counts, and capture its exact variantSetDbId (MODULE§project§run) — every step below needs it.
3. QC sweep (all take the variant_set_db_id from step 2):
   • qc_call_rate(variant_set_db_id) — worst samples/markers.
   • qc_heterozygosity(variant_set_db_id) — outliers (heed the high-Ho warning: inflated heterozygosity often means contamination or a miscoded import).
   • qc_maf_filter(variant_set_db_id) — how many markers a MAF/missingness filter drops.
   • qc_duplicate_accessions(variant_set_db_id) — clones / mislabelled duplicates.
4. Audit with audit_import_quality(variant_set_db_id) — catches genotype-encoding artifacts (e.g. everything called, no heterozygotes, suspicious monomorphic fraction).
5. Summarise flagged samples/markers and give a clear verdict on whether the import looks clean.