HaploSOFT
Enter the known alleles (allelic or serological notation) and click Predict.
Reference base : Fr Gr
Free-text input — type here, alleles are extracted automatically
Parsed alleles (read-only preview)
ABC
Allele 1 Allele 2 Allele 1 Allele 2 Allele 1 Allele 2
DRB1DQB1DQA1DRB345DPA1DPB1
Allele 1 Allele 2 Allele 1 Allele 2 Allele 1 Allele 2 Allele 1 Allele 2 Allele 1 Allele 2 Allele 1 Allele 2

Examples: A*01:01 or A1 · B*08:01 or B8 · Cw7 · DRB1*03:01 or DR17 · DQ2 · Nop

⚠ Minimum required: A + B + DRB1 for ABC group · DRB1 + DQB1 for class II groups

ABC-DRB1-DQB1
DRB1-DQA1-DQB1
DRB1-DQB1-DRB345
DPA1-DPB1
Run a prediction to see results for ABC-DRB1-DQB1
Run a prediction to see results for DRB1-DQA1-DQB1
Run a prediction to see results for DRB1-DQB1-DRB345
Run a prediction to see results for DPA1-DPB1
This tool is dedicated to research use only. It has not been validated for clinical use. Using the data provided for clinical purposes is under the sole responsibility of the user.
⚡ Batch & Export
📂 Batch prediction Excel input → Excel output (Marginal · Diplotype · Inputs)
↓ Template
💾 Export current prediction 3 sheets: Marginal · Diplotype · Tables per group
Exports all predicted groups with cascade-corrected alleles.
⚠ Do not enter any patient-identifying information (name, date of birth, medical record number, etc.) — including in batch mode: only HLA typing data should be used. No prediction is stored on the server — export your results (Excel) before leaving this session, as nothing can be recovered afterwards.
Developed by Valentin Clichet, biologist at the Transplantation Laboratory, Saint-Louis Hospital (Paris). For any question, please contact the developer directly.
📊 Database information
Statistics computed from the reference haplotype databases used for prediction.
🇫🇷 France (Fr)
🇬🇷 Greece (Gr)
Loading…
Loading…
🔍 Search an allele in the database
Type an allele name to search.
⚠ Known G-group ambiguities — Greece model
The Greek database was built using G-group allele calls. The following alleles were unknown at the time of training and were merged with the listed allele. Predictions involving these alleles may be ambiguous.
Loading…
ⓘ HaploSOFT — User Guide
Click any section to expand it. Examples use: A*02/A*03 · B*35/B*49 · C*04/C*07 · DRB1*01/DRB1*04 · DQB1*05/DQB1*03:01.
✎1 — Entering alleles▶

Two input modes: Free-text — paste any text, alleles extracted automatically. Manual grid — enter locus by locus.

SerologicalAntigen group only — e.g. A2, B35, DR4. Most likely allele within that group will be predicted. First-fieldFirst field only — e.g. A*02, DRB1*04. Equivalent to serological in allelic format. Full allelicComplete allele — e.g. A*02:01, DRB1*04:05. Most precise input. MixedMix notations across loci — e.g. A*02:01 for A and DR4 for DRB1.
Example
A2 → sero  |  A*02 → first field  |  A*02:01 → full allelic
Free-text: A*02 A*03 B*35 B*49 C*04 C*07 DRB1*01 DRB1*04 DQB1*05 DQB1*03:01 → grid fills automatically.

Leave empty = unknown (predicted). Nop = DRB345 explicitly absent. Min required: A + B + DRB1.

🆕2 — "Predicted complete haplotypes" banner▶

Most probable complete diplotype assembled from each locus group's top prediction.

Example
Hap. 1: A*03:01 · B*35:01 · C*04:01 · DRB1*01:01 · DQB1*05:01 · DRB345 Absent · DPA1*01:03 · DPB1*04:01
Hap. 2: A*02:01 · B*49:01 · C*07:01 · DRB1*04:05 · DQB1*03:01 · DRB4*01:03 · DPA1*01:03 · DPB1*02:01
marginal conditionalEach allele chosen from its marginal probability across all compatible diplotypes. top diplotypeAlleles taken directly from the rank-1 diplotype. ⚠ ReconstructedNo compatible diplotype found — alleles reconstructed haplotype by haplotype, shown in orange.
📊3 — Compatible haplotypes table▶

All diplotypes compatible with the input, ranked by probability. Header: total compatible / database total + cumul. top 5.

31 413
Haplotypes in database
All distinct haplotypes for this group
⟶matches at least one of your two alleles, per locus
86
Compatible haplotypes
Individually compatible — not yet paired
⟶paired so both alleles are covered on every locus at once
≤ 25
Valid diplotype pairs
Shown in the table below, ranked by probability (Top 1: 59.9%)

Counter-intuitive case: the last step can be much smaller than the middle one — e.g. below, 4 individually-compatible haplotypes yield only 2 valid pairs, because one of them can't cover both your alleles together with any of the other three.

▶Show a real worked example — compatible list vs. valid pairs

Real case from the Fr base (group ABC-DRB1-DQB1), using serological input — as you normally would. Input: A25 / A31 at locus A and B78 / B81 at locus B.

4 compatible haplotypes (match at least one input serotype per locus, independently)
H1 A25~B78~Cw16~DQ6~DR15 — matches A25, B78 (A*25:01~B*78:01…)
H2 A31~B78~Cw16~DQ6~DR13 — matches A31, B78 (A*31:01~B*78:01…~DRB1*13:02)
H3 A25~B81~Cw18~DQ7~DR11 — matches A25, B81 (A*25:01~B*81:01…)
H4 A31~B78~Cw16~DQ6~DR13 — matches A31, B78 (A*31:01~B*78:01…~DRB1*13:01)

Only H3 carries B81, so every valid pair must include H3 — and H3 carries A25, so its partner must carry A31 to cover both serotypes at locus A: that's H2 and H4. Pairing H3 with H1 would leave A31 uncovered by either haplotype, so it doesn't count as a valid diplotype even though H1 is individually compatible.

2 valid diplotype pairs
H2 × H3 → 50.0%
H4 × H3 → 50.0%

H2 and H4 look identical in serological notation (A31~B78~Cw16~DQ6~DR13) but are different alleles at DRB1 (*13:02 vs *13:01) — hence 2 separate valid diplotypes, both with H3. H1 never appears in the table below: it's counted in the "4 compatible" total, but it doesn't validly pair with any of the other 3.

#Rank of the diplotype. Haplotype 1/2 (sero)Serological notation. In parentheses: allelic notation for DRB1-DQB1-DRB345. Prob. %Normalised probability among all compatible diplotypes.
Example — 86 compatible / 31 413 total · Top 1: 59.9% · Cumul. Top 5: 80.0%
#1 A*03:01-B*35:01 × A*02:01-B*49:01-DRB1*04:05 → 59.86%
#2 same Hap.1 × A*02:05-B*49:01-… → 8.16%

Colours: ≥10% ≥5% ≥1% <1% — Click a row for full detail in the right panel.

🔍4 — Right panel: Selected diplotype▶
Haplotype freq.Frequency, rank and carrier count in the database. E.g. Freq. 0.008178, rank 6/31 413, N=1 863. First-field freq.Frequency of this first-field allele among all first-field alleles. E.g. A*03:01 → 98.241% (rank 3/16). Locus freq.Frequency of this specific allele in the full reference population. E.g. A*03:01 → 11.641% (rank 3/140). Indiv. freq.Estimated carrier frequency. E.g. A*03:01 → 21.811%, N=35 387. Assigned / CIWDSerological antigen (e.g. A3) and CIWD class: C=Common, I=Intermediate, W=Well-documented, D=Documented. Top B / Top CMost frequent alleles at another locus on haplotypes carrying this allele, with conditional frequencies. Click ↓ Excel for the full contingency table.
📈5 — Right panel: Marginal probabilities per allele▶

Probability of each allele normalised within its serological antigen group across all compatible diplotypes. For example with input A2: among all compatible haplotypes in the A2 group, 88.7% carry A*02:01 — weighted by allele frequency across all known haplotypes matching the input. Click a bar to highlight diplotypes containing that allele.

Example — Locus A & DRB1
— A3 — (among A3-group compatible haplotypes)
A*03:01
99.0%
A*03:02
1.0%
— A2 — (among A2-group compatible haplotypes)
A*02:01
88.7%
A*02:05
11.4%
A*02:02
1.9%
— DR4 —
DRB1*04:05
82.2%
DRB1*04:01
10.4%
DRB1*04:07
4.6%

Bar colours: ■ ≥50%  ■ ≥10%  ■ ≥1%  ■ <1%

🌪6 — Right panel: Sankey diagram▶

Flow diagram A → B → C → DRB1 → DQB1. Each ribbon is a haplotype; width ∝ probability. Multiple ribbons from one allele node show alternative paths (e.g. A*03:01 → B*35:01 at 93% vs B*18:01 at 5%).

Example — ABC-DRB1-DQB1
ABCDRB1DQB1 A*03:01B*35:01B*18:01C*04:01DRB1*01:01*01:02DQB1*05:01 A*02:01B*49:01B*44:02C*07:01DRB1*04:05*04:01DQB1*03:01 ~60%~40% 93%5% 88%7%
A*03:01 haplotypes (teal) — B*35:01 93%, B*18:01 5% A*02:01 haplotypes (purple) — B*49:01 88%, B*44:02 7%
HoverShows haplotype label, frequency, rank and N= carriers. Highlight selectedHighlights the selected diplotype's two haplotypes in teal and purple. Group by inputReorders and colours allele nodes by their true serological group (e.g. B15 → B62/B63 shown together), based on the reference database's own broad/split association — not just the exact text you typed. Nodes matching your first vs second input allele are tinted teal/purple. Group tabsSankey reflects the active group tab. Switch to see DRB1-DQA1-DQB1 etc.
🔖7 — Group tabs▶
ABC-DRB1-DQB1Main class I+II group. Requires A+B+DRB1. Example: 86/31 413. DRB1-DQA1-DQB1Class II, predicts DQA1 if unknown. Example: 15 compatible. DRB1-DQB1-DRB345Predicts DRB3/4/5 from DRB1. DR1, DR8, DR10 → always Absent. Example: 10. DPA1-DPB1Class II DP, predicts DPA1 if unknown. Example: 30.
⚛8 — Cascade prediction▶
Example
Input DRB1*01 → ABC group predicts DRB1*01:01 (71.5%)
→ DRB1-DQB1-DRB345 receives DRB1*01:01 → DRB345 Absent (DR1 has no DRB3/4/5)
⚡9 — Batch mode▶
TemplateDownload for required columns: ID, A1, A2, B1, B2, C1, C2, DRB11, DRB12, DQB11, DQB12, DRB3451, DRB3452, DPA11, DPA12, DPB11, DPB12. Output — 3 sheetsMarginal — the reference prediction, same principle as the "Predicted complete haplotypes" banner above (marginal probability per locus, not just the single best diplotype). Diplotype — the strict rank-1 diplotype per locus group, which can occasionally differ from Marginal on ambiguous loci. Inputs — your original alleles as parsed from the file, for traceability. Orange cellsSame meaning as "⚠ Reconstructed" in manual mode: no compatible diplotype existed, that allele was reconstructed rather than matched directly. Progress barReal-time indicator: current patient / total, with a fill percentage.

Batch and manual mode now use exactly the same cascade and marginal-probability logic — the same input gives the same prediction either way.