Skip to content

HLA (imputed)

Not a medical test — for research and educational use only

Yeliztli analyses consumer genotyping-array data (23andMe / AncestryDNA), which is not a clinical-grade test. Results are not diagnostic, are not clinically validated, and must not be used to make medical decisions. Array data is especially unreliable for rare, disease-causing variants. Always confirm any finding with an accredited clinical laboratory and discuss it with a qualified clinician or genetic counsellor before acting on it. See the Intended use & disclaimers page for the details and the evidence behind this warning.

Provisioned feature

The HLA (imputed) page is visible in the app for every install, but it stays empty until an operator installs the HIBAG runtime, supplies ancestry-specific HIBAG model files, and runs the HLA prediction script for a sample. An empty page usually means "not provisioned yet", not a broken sample.

HLA (imputed) predicts classical HLA alleles from SNP-array data and uses those calls to power four surfaces:

  • Drug hypersensitivity — HLA-drug safety findings for abacavir, carbamazepine, oxcarbazepine, phenytoin/fosphenytoin, allopurinol, and dapsone [1-6,9-13].
  • Disease rule-outs — celiac HLA-DQ and narcolepsy type 1 HLA-DQB1*06:02 context.
  • Autoimmune susceptibility — selected HLA susceptibility markers such as HLA-B*27, HLA-C*06:02, rheumatoid-arthritis shared epitope alleles, and type-1-diabetes DR-DQ patterns.
  • Raw imputed-HLA viewer/export — the called HLA loci, posterior probability, low-confidence flag, source, and ancestry model.

HIBAG is an HLA imputation method: it predicts HLA alleles from dense SNP genotypes using attribute bagging, with published validation performance that depends on locus, array content, and ancestry [7]. Later comparisons show that SNP-based HLA imputation is still not as accurate as molecular HLA genotyping, especially for harder loci and diverse populations [8].

How to read it

Treat every result as a screening lead, not a clinical HLA type.

  • Imputed is not typed. Confirm any result that could change medication, diagnosis, or care with clinical high-resolution HLA typing.
  • Never use imputed HLA for transplant, organ, or stem-cell donor/recipient matching. The raw viewer repeats this guard and embeds it in exported CSV files.
  • A positive drug-safety finding is not a medication order. It means the imputed allele matches an HLA-drug association that should be discussed with a clinician and confirmed before changing therapy [1-6,9-13].
  • A negative finding is not universal reassurance. HLA-drug test performance varies by drug pair and population; many pairs have high negative predictive value but low positive predictive value [9].
  • Ancestry and model choice matter. Use the model that best matches the sample ancestry, and interpret mismatches or admixed ancestry with extra caution.
  • Low-confidence calls are flagged, not hidden. They remain visible so you can see why a section is uncertain. Drug-safety rows with low-confidence imputed calls are indeterminate: they should not be treated as reliable positives or negatives until confirmed with clinical high-resolution HLA typing.

Why the page may be empty

Default Yeliztli installs do not run HLA imputation automatically. If a sample has no persisted hla_calls table entries, the HLA page reports that no imputed HLA calls are available. The Allergy module's single-tag HLA proxy fallback may still produce limited HLA-proxy findings, but that is separate from this first-class imputed-HLA page.

Operator setup

HIBAG is intentionally a bring-your-own runtime and model workflow. Yeliztli does not bundle the R/Bioconductor HIBAG package or pre-fit HLA model files.

  1. Install R, Rscript, and the Bioconductor HIBAG package on the machine that runs the backend.
  2. Fetch the required pre-fit HIBAG model files yourself, respecting their license terms. Yeliztli expects files named {ancestry}-HLA4.RData; supported ancestry names are European, Asian, Hispanic, and African.
  3. Configure the runtime and model directory:
export YELIZTLI_HIBAG_RSCRIPT=/usr/bin/Rscript
export YELIZTLI_HIBAG_MODEL_DIR=/path/to/hibag-models

The same values can be set as hibag_rscript and hibag_model_dir in ~/.yeliztli/config.toml.

  1. Prepare the sample's HLA-region PLINK input from an already annotated sample database:
python scripts/prepare_hla_input.py \
  --sample-db ~/.yeliztli/samples/<sample>.db \
  --out-prefix ~/.yeliztli/hla/<sample>/<sample>
  1. Run prediction and persist the calls:
python scripts/predict_hla.py \
  --sample-db ~/.yeliztli/samples/<sample>.db \
  --work-dir ~/.yeliztli/hla/<sample> \
  --ancestry European
  1. Reopen HLA (imputed) for that sample. You can also check /api/hla/status to confirm whether Rscript and at least one ancestry model are visible to the backend.

Model files are not Yeliztli reference data

HIBAG model files are user-supplied external inputs, not release-bundled Yeliztli reference data. See the external inputs strategy for the license posture.

References

[1] Martin MA, Hoffman JM, Freimuth RR, et al. Clinical Pharmacogenetics Implementation Consortium Guidelines for HLA-B Genotype and Abacavir Dosing: 2014 Update. Clinical Pharmacology & Therapeutics. 2014. DOI: 10.1038/clpt.2014.38; PMID: 24561393.

[2] Leckband SG, Kelsoe JR, Dunnenberger HM, et al. Clinical Pharmacogenetics Implementation Consortium Guideline for HLA Genotype and Use of Carbamazepine and Oxcarbazepine: 2017 Update. Clinical Pharmacology & Therapeutics. 2018. DOI: 10.1002/cpt.1004; PMID: 29392710.

[3] Karnes JH, Rettie AE, Somogyi AA, et al. Clinical Pharmacogenetics Implementation Consortium (CPIC) Guideline for CYP2C9 and HLA-B Genotypes and Phenytoin Dosing: 2020 Update. Clinical Pharmacology & Therapeutics. 2021. DOI: 10.1002/cpt.2008; PMID: 32779747.

[4] Hershfield MS, Callaghan JT, Tassaneeyakul W, et al. Clinical Pharmacogenetics Implementation Consortium Guidelines for Human Leukocyte Antigen-B Genotype and Allopurinol Dosing. Clinical Pharmacology & Therapeutics. 2013. DOI: 10.1038/clpt.2012.209; PMID: 23232549.

[5] Zhang FR, Liu H, Irwanto A, et al. HLA-B*13:01 and the Dapsone Hypersensitivity Syndrome. New England Journal of Medicine. 2013. DOI: 10.1056/NEJMoa1213096; PMID: 24152261.

[6] Liu H, Wang Z, Bao F, et al. Evaluation of Prospective HLA-B*13:01 Screening to Prevent Dapsone Hypersensitivity Syndrome in Patients With Leprosy. JAMA Dermatology. 2019. DOI: 10.1001/jamadermatol.2018.5360; PMID: 30916737.

[7] Zheng X, Shen J, Cox C, et al. HIBAG - HLA Genotype Imputation with Attribute Bagging. The Pharmacogenomics Journal. 2014;14:192-200. DOI: 10.1038/tpj.2013.18.

[8] Pappas DJ, Lizee A, Paunic V, et al. Significant variation between SNP-based HLA imputations in diverse populations: the last mile is the hardest. The Pharmacogenomics Journal. 2018;18:367-376. DOI: 10.1038/tpj.2017.7.

[9] Manson LEN, Swen JJ, Guchelaar HJ. Diagnostic Test Criteria for HLA Genotyping to Prevent Drug Hypersensitivity Reactions: A Systematic Review of Actionable HLA Recommendations in CPIC and DPWG Guidelines. Frontiers in Pharmacology. 2020;11:567048. DOI: 10.3389/fphar.2020.567048.

[10] Kaniwa N, Saito Y, Aihara M, et al. HLA-B*1511 is a risk factor for carbamazepine-induced Stevens-Johnson syndrome and toxic epidermal necrolysis in Japanese patients. Epilepsia. 2010;51:2461-2465. DOI: 10.1111/j.1528-1167.2010.02766.x; PMID: 21204807.

[11] Wong CSM, Yap DYH, Ip P, et al. HLA-B*15:11 status and carbamazepine-induced severe cutaneous adverse drug reactions in HLA-B*15:02 negative Chinese. International Journal of Dermatology. 2022;61:486-491. DOI: 10.1111/ijd.15792; PMID: 34553372.

[12] Biswas M, Ershadian M, Shobana J, Nguyen AH, Sukasem C. Associations of HLA genetic variants with carbamazepine-induced cutaneous adverse drug reactions: An updated meta-analysis. Clinical and Translational Science. 2022;15:1887-1905. DOI: 10.1111/cts.13291; PMID: 35599240.

[13] Manson LEN, Nijenhuis M, Soree B, et al. Dutch Pharmacogenetics Working Group (DPWG) guideline for the gene-drug interaction of CYP2C9, HLA-A and HLA-B with anti-epileptic drugs. European Journal of Human Genetics. 2024;32:903-911. DOI: 10.1038/s41431-024-01572-4; PMID: 38570725.