One platform for HLA, KIR & LILR immunogenetics.
TRIAD is a validated, local-first analysis platform for case-control association and population genetics across the three gene systems that drive innate and adaptive immune variation — no assembling separate tools, no data leaving your machine.
Two study designs, one genotype table.
A typical immunogenetic study touches allele frequencies, association testing, amino-acid analysis, haplotypes, and diversity statistics — usually across several separately-validated tools, with data reformatted by hand at every handoff. TRIAD imports one standardized table and runs either track, or both, without switching tools.
Association testing
Find and quantify disease associations, allele- and amino-acid-level.
- Allele & carrier-model association
- Amino-acid association (BIGDAWG-style)
- Haplotype association
- Meta-analysis (fixed & random effects)
- HED (heterozygosity/divergence) scoring
Descriptive & structural analysis
Characterize a cohort or compare populations in depth.
- Allele & haplotype frequencies (EM)
- Linkage disequilibrium (D′, Wn, ALD)
- Hardy-Weinberg equilibrium
- Fst / AMOVA, gene diversity
- Phylogeny (NJ / UPGMA)
Functional interactions
The analysis most tools don't offer at all.
- HLA–KIR receptor–ligand pairing
- Missing-ligand analysis
- HLA–LILR & KIR–LILR (LRC) linkage
- KIR / LILR genotype & CNV calling
- Gene-content carrier association
HLA, KIR, and LILR, analyzed together, not adjacently. Most tools handle one gene system well. TRIAD's functional-interaction layer treats receptor–ligand relationships across all three as first-class analyses, not a manual cross-reference step at the end.
Every method checked against an external reference.
Not "tested" in the informal sense — each analysis is reproduced against a named gold standard (BIGDAWG, Arlequin, statsmodels, or the primary literature) and the concordance is recorded, re-runnable from the repository's validation/ suite.
| Method | Reference / gold standard | Concordance |
|---|---|---|
| Allele frequencies | BIGDAWG (Pappas 2016) | exact, max diff 0 |
| Case-control association | BIGDAWG (Pappas 2016) | OR exact |
| Hardy-Weinberg equilibrium | BIGDAWG (Pappas 2016) | exact (χ², dof, p) |
| AMOVA / Fst | Arlequin 3.5.2.2 | exact vs. Arlequin |
| HED (class I) | Pierini & Lenz (2018) | exact to 1e-9 |
Show all 21 validated methods
| Amino-acid association | BIGDAWG (Pappas 2016) | omnibus χ² exact |
| Conditional amino-acid | statsmodels (logistic LRT) | match to ~1e-4 |
| EM haplotype frequencies | Excoffier & Slatkin (1995) | known-answer, max err ~5e-3 |
| Haplotype association | scipy Fisher + BIGDAWG (OR) | Fisher exact match |
| Gene diversity & heterozygosity | Arlequin 3.5.2.2 / Nei (1987) | exact vs. Arlequin |
| Ewens-Watterson neutrality | Ewens (1972) / Slatkin (1996) | reproduces analytic moments |
| Linkage disequilibrium (D′, Wn, ALD) | Hedrick (1987) / Thomson-Single (2014) | matches worked example, 7 dp |
| Logistic regression (Wald) | statsmodels | match to ~1e-6 |
| Logistic regression (Firth) | independent penalised-likelihood optimiser | match to ~1e-8 |
| Meta-analysis (DerSimonian-Laird) | statsmodels (DL) | match to ~1e-4 |
| Meta-analysis (Mantel-Haenszel) | Robins-Breslow-Greenland (1986) | OR + variance exact |
| Functional KIR-ligand pairing | Immunogenetics literature | matches published rules |
| Phylogeny (NJ / UPGMA) | Biopython DistanceTreeConstructor | exact topology match |
| KIR / LILR gene-content association | scipy Fisher + BIGDAWG | Fisher exact |
| HED (class II) | no external gold standard | internal logic verified |
Full methodology and reproduction scripts live in the repository's validation/ suite.
Load, analyze, report.
The three-step workflow TRIAD is built around, shown on the bundled demo cohort.
Your genotype data never leaves your machine.
TRIAD is a desktop application, not a web upload service. Analysis runs locally, which matters for patient-derived data and sidesteps the data-governance questions a cloud tool would raise.
Request access.
TRIAD isn't publicly listed yet — the repository is private ahead of publication. Tell us a little about yourself and confirm your email, and you'll be taken straight to the download page.