Reproduce ClinVarbitration decisions from archived ClinVar flat files — rclinvarbitration_reproduce_clinvarbitration

Reproduces Centre for Population Genomics ClinVarbitration 2.2.11's allele-level decision algorithm directly from NCBI's versioned submission_summary and variant_summary files. The generated Parquet has the exact seven-column clinvar_decisions.tsv layout: contig, position, reference, alternate, clinical_significance, gold_stars, and allele_id.

Usage

rclinvarbitration_reproduce_clinvarbitration_parquet(
  con,
  submission_path,
  variant_path,
  path,
  assembly = c("GRCh38", "GRCh37"),
  submitter_exclusions = character()
)

Arguments

con: A DuckDB DBI connection.
submission_path: Archived NCBI submission_summary_YYYY-MM.txt.gz.
variant_path: Archived NCBI variant_summary_YYYY-MM.txt.gz.
path: New output .parquet path.
assembly: Genome assembly: "GRCh38" or "GRCh37".
submitter_exclusions: Submitter names to exclude for a blinded run.

Value

A named list describing the written Parquet file, invisibly.

Details

This is separate from rclinvarbitration_import_xml() and the disease-aware policy views. It exists to reproduce and differentially validate the upstream allele-level artifact on matching archived flat-file releases. It uses DuckDB's streaming CSV reader and SQL aggregation; it does not invoke Python, Hail, VEP, or PM5 logic.