For Research Use Only · Not for use in diagnostic procedures.
A deterministic breast-cancer digital twin

One patient. One continuous record. One living model.

OncoBreast21 is a transparent, deterministic digital twin for breast cancer — the published 21-gene recurrence score at its core, extended across the disease journey into one auditable, longitudinal model. No machine learning in the score; externally validated across nine independent public cohorts. Built for researchers and clinicians who need an auditable engine, not a black box.

Nine cohorts · Cohen’s κ 0.94 · C-index matched to reference — reproducible & auditable.

For Research Use Only — not a diagnostic device. Intended for research, not for guiding patient treatment.

Expression Scoring 21-gene recurrence-risk score Hereditary Mutations Germline BRCA1/2 & panel genes Pharmacogenomics Drug–gene safety & dosing Signals · one record
21-gene recurrence-score scale one clear result
Lower riskHigher risk
Low Intermediate High
Breast cancer awareness — a woman wearing a pink ribbon
Interactive Journey

Your Digital Twin, Step by Step

From inherited risk through diagnosis, genomic prognosis, therapy and follow-up — how a patient's course is modelled as one continuous digital twin, each stage reading from and writing back to the same record.

Inherited-risk context

Germline findings (BRCA1/2, PALB2 and related genes), when available, are recorded in the twin to contextualise the recurrence signal. Research module — not a germline testing service.

BRCA1 · BRCA2 · PALB2 · CHEK2 · panel genes

Prevention

Confirm the diagnosis, profile the tumour

ER / PR / HER2, nodal status and grade are captured alongside the genomic signal so the recurrence score is interpreted in its clinical context.

Imaging · Biopsy · ER / PR / HER2 status

Diagnosis

One clear recurrence-risk score

For ER-positive, HER2-negative early breast cancer, the deterministic 21-gene module places distant-recurrence risk on a Low–High scale — the validated core of the platform.

21-gene expression score · Low → High

Prognosis Low High

Pharmacogenomic context

Where pharmacogenomic results (e.g. CYP2D6, DPYD) are available, they are recorded in the twin as research context for the treating team. The platform does not issue dosing instructions.

CYP2D6 · DPYD · drug–gene dosing

Treatment

Surveillance context

Longitudinal signals (e.g. ctDNA, imaging) can be recorded over time as research context. The platform does not check prescriptions or issue clinical alerts.

Interaction checks · structured surveillance

Follow-up
Evidence

Reproducible, and externally validated

OncoBreast21 reproduces the published 21-gene recurrence score deterministically. Against an independent reference implementation across nine public cohorts (~5,488 patients): near-perfect agreement, and a Harrell C-index that matches the reference in every cohort with a survival endpoint.

Agreement

Cohen’s κ 0.94

Risk-band assignment agrees with an independent implementation of the published formula on the same patients (Spearman up to 0.998).

Discrimination

C-index matched

Harrell C-index matches the reference implementation in every cohort carrying a survival endpoint.

Transparency

Deterministic · no ML

Version-locked and content-hash-audited; recalibratable to a local population. Methods and manuscript available on request.

Method

A transparent, deterministic engine

The published 21-gene methodology, computed exactly and reproducibly — every number is traceable to its inputs.

Reference normalisation

Expression is normalised to the reference genes exactly as specified; QC gates fail unusable inputs safely.

Group scores

Proliferation, invasion, HER2 and ER group scores are combined using the published coefficients.

Deterministic score & band

A single unscaled value maps through the fixed scale to a Low / Intermediate / High band — no machine learning.

Version-locked & auditable

The configuration is content-hash-stamped and drift-guarded, so a given input always yields the same result.

Intended use

Where the engine applies

For Research Use Only. The recurrence-score module addresses one clearly-defined population.

✓ In scope

  • Hormone-receptor-positive (ER-positive)
  • HER2-negative
  • Early-stage, invasive
  • Node-negative, or 1–3 positive nodes

✗ Out of scope

  • HER2-positive disease
  • Triple-negative disease
  • Non-invasive changes (DCIS / LCIS only)
  • Advanced, inflammatory, or male breast cancer
Approach & status

Anchored to a method used worldwide

OncoBreast21 reconstructs the 21-gene methodology studied in large clinical trials (TAILORx, RxPONDER). External validation to date is on public array and RNA-seq cohorts; an independent clinical-cohort validation (FFPE RT-qPCR) is planned.

Method

Established & published

A well-studied way of reading tumour genes to estimate distant-recurrence risk.

Human-in-the-loop

Decision-support only

Advisory to the clinician and tumour board — it does not diagnose or decide.

Open

Evidence on request

Validation summary, methods and manuscript draft available to researchers and clinicians.

What OncoBreast21 is — and isn’t

Stated plainly

OncoBreast21 is a transparent reconstruction of the 21-gene recurrence score, for research use. It is not a diagnostic device, it does not make treatment decisions, and it is not CE-marked or FDA-cleared.

Research Use Only. Not for use in diagnostic procedures; the treating clinician makes every decision.
Grounded in established science. The recurrence-score module reconstructs a method studied in large clinical trials (TAILORx, RxPONDER).
Externally validated. Nine public cohorts; an independent clinical-cohort (FFPE RT-qPCR) validation is planned.
Transparent by design. Deterministic, version-locked, content-hash-audited — no machine learning in the score.

For researchers & clinicians

Request the validation summary and methods, or discuss a research collaboration.

Contact the team