The graduation ladder
Describe the place
Record tissue architecture, compartments, species, assay conditions, published measurements and disagreements. Preserve missing values as missing—not zero.
Identify the causal link
Ask which measurable quantity affects oxygen, nutrients, immune activity or actual drug exposure. A proxy like vessel density does not equal intracellular exposure.
Specify the mapping
Define units, reference range, normalization, uncertainty, applicability and failure cases. A valid mathematical transform is necessary but not sufficient.
Challenge and calibrate
Compare predictions with independent measurements, not only the data used to tune the model. Record where the model fails. A user-entered validation claim still requires review.
Run a versioned test chamber
Only a reviewed, separately implemented model earns an executable named-environment preset. Keep the source dataset, version and uncertainty with every experiment.
Three environments. Three different problems.
Getting there may be the hard part.
Dense stroma and compressed vessels suggest a transport problem. But murine pressure estimates cannot simply become human drug-exposure values. The needed bridge is measured delivery and a validated transport model.
A barrier is not a lower dose.
Blood–brain and blood–tumor barriers constrain passage differently by molecule type and location. Antibody access, small-molecule permeability and tumor heterogeneity require separate measurements; one “brain penetration” dial hides that biology.
Which side of the cell can you reach?
A target can be abundant yet poorly accessible from the vascular side because of epithelial polarity. Its position matters as much as its expression; the model needs compartment and orientation, not just a target count.
What a result can teach us
While named environments are being calibrated, our six mechanistic role panels deliberately stress one question at a time: delivery, cycling, repair, apoptosis execution, immune pressure or permissive reference conditions.
- Compare the candidate with its parent under identical conditions.
- Track mechanism-specific injury separately from cell fate.
- Keep hypotheses about improved chemistry separate from demonstrated properties.
- Ask where the advantage disappears and which measurement could falsify the explanation.
- Do not equate a higher normalized injury score with clinical benefit or safety.
Research is the bridge.
The microenvironment workspace now shows a research-readiness ledger for each selected Atlas environment. It reports whether evidence is only indirect, whether a mapping is specified, and what must happen before a model upgrade. None of those labels automatically changes the simulator.