bmyCure4MM

bmyCure4MM is an E1 research prototype whose north-star is a virtual and reproducible Multiple Myeloma research laboratory.

intended_use_level: E1_research_prototype
clinical_decision_support: false
patient_specific_prediction_validated: false
causal_effect_identified: false

The platform currently provides governed structured research data, lineage-bound Patient Twin infrastructure, mechanistic PK/PD and schedule-aware simulation, model-relative counterfactual research, immutable run identity, artifact integrity and scientific QA foundations.

It does not choose treatment, issue patient-specific dose or schedule instructions, predict validated patient benefit or identify causal treatment effects.

Start here

• Mission and north star
• Current verified state
• Capabilities and limitations
• Canonical intended use
• Source-of-truth policy
• Current architecture
• Target virtual laboratory
• Current model registry and formulas
• Project roadmap

Current verified baseline

repository = andreazedda/bmyCure4MM
branch = master
verified_head = a33418fb8ae9cb9fd05832dd9bc1cb0778e08533

Current M0-R path:

#14 documentation/source of truth
→ #13 mandatory CI aggregate gate and PR governance
→ #23 minimum scientific/privacy invariants
→ #26 close M0-R

The canonical dependency baseline now uses Django 5.2.17 LTS and sqlparse 0.6.0. Dependency audit, Django compatibility/deploy checks, disposable migrations, Python 3.11/3.12 suites, numerical identity, Secret Scan, Docker build and the protected required gate passed in PRs #71 and #72.

Current capabilities

• versioned Structured Research Dataset Contract and idempotent import infrastructure;
• content-addressed computational-input and Twin lineage;
• persistent Twin states, residuals and calibration infrastructure;
• logistic tumour/healthy-cell and PK/PD research simulation;
• day-resolved administered-dose profiles and schedule identity;
• heuristic hepatic and neutropenia risk signals;
• mechanistic baseline/alternative counterfactual runs;
• immutable run manifests, model registry, artifact hashes, comparability and invalidation;
• uncertainty, sensitivity, robustness and backtesting prototypes;
• optional molecular research utilities;
• separate learning, research and administration surfaces.

See Capabilities and limitations before interpreting any result.

Local setup

Python 3.11 and 3.12 are supported. uv.lock is the sole dependency lock and uv 0.12.3 is required.

git clone https://github.com/andreazedda/bmyCure4MM.git
cd bmyCure4MM
python3.11 -m pip install uv==0.12.3
uv sync --frozen --extra chemistry
uv run python manage.py migrate
uv run python manage.py check
uv run python manage.py runserver

Omit --extra chemistry for core work that does not require RDKit. Dependency operations are documented in Dependency Operations.

Verification and safety

uv lock --check
uv sync --frozen --extra chemistry
uv run python manage.py check
uv run python manage.py makemigrations --check --dry-run
uv run python manage.py test
uv run ruff check .
uv run ruff format --check .
uv run mypy
uv run python -m scripts.check_repository_hygiene
uv run python -m scripts.check_numerical_baseline
uv run python -m scripts.audit_dependencies
bash scripts/pre_push_research_safety_check.sh

Current test boundaries and CI limitations are documented in Testing and Dependency Audit Triage.

Privacy

Names, medical-record numbers, dates of birth, clinical PDFs, source excerpts, direct identifiers, private dataset payloads and private artifacts must never enter Git. Keep private research material under ignored local_private/ paths and export only explicitly validated, de-identified research artifacts.

Documentation status

The documentation corpus is governed by:

• Documentation policy
• Documentation inventory

Historical feature and release narratives are retained only as archives and are not current product claims.

License and security

See LICENSE and SECURITY.md.