Skip to content

Assumptions and Limitations

Research use only

genoABO is intended for research use only. Genotype-inferred ABO groups do not capture all rare, weak, cis-AB, structural, or otherwise atypical ABO alleles and must not be used for transfusion or clinical decision-making.

Scope

  • Common A/B/O/AB inference from four SNP markers
  • Six common diplotypes under standard assumptions
  • Extended method limited to ABO*A.01, ABO*B.01, ABO*O.01 (allele-registry v0.1)
  • No ISBT-comprehensive typing
  • No structural variant detection
  • No clinical validation claims

Known limitations

  1. Rare alleles: Non-deletional O alleles, weak A/B, cis-AB, and null alleles may be misclassified.
  2. Proxy inference: rs505922 is in LD with O, not functional; UKB fallback lowers confidence. 1000 Genomes 30x hg38 haplotype r² with rs8176719: EAS 0.946, EUR 0.869, AMR 0.820, AFR 0.437. See Marker LD.
  3. Reference orientation: rs8176719 representation varies by pipeline; normalization required.
  4. Population assumptions: The default --assume-bo rule (O-del on A background → B/O) is a common-allele assumption, not phase. On 1000 Genomes NYGC 30x GRCh38 (phased GT, not serology), unphased double-hets were B/O in AMR 20/20, EAS 124/124, and EUR 41/41. In AFR 105/113 (92.9%) were B/O and 8/113 were A/O, so --assume-bo would call group B instead of A. Pass --no-assume-bo to leave unphased double-hets ambiguous (recommended for unphased African-ancestry data). See 1000 Genomes --assume-bo check.
  5. Imputation quality: The VCF INFO score is shown in --inspect / .qc.tsv when present. It is not used to change confidence. Dosage or GP below the hardcall thresholds is treated as a missing marker (confidence = none).

Unresolved edge cases (TODO)

  • Comprehensive validation against serology across ancestries
  • Non-deletional O subtypes (rs41302905, rs782782485 region)
  • Multiallelic ABO copy-number or hybrid alleles
  • CHM13 coordinate validation in repetitive 9q34.2 region
  • Strand ambiguity for palindromic contexts (if any)

Population frequency QC

Population A/B/AB/O percentages in QC output are informational only. Samples are never rejected based on population frequency deviation.

Build detection

When evidence conflicts, genoABO errors and requires explicit --build. Never silently guesses assembly.