Virtual Cell Capability Atlas

See exactly what NMD-VCell can represent, predict and not yet claim.

Five capability levels connect evidence organization to future disease prediction. Every level names its input, output, evaluation and unlocking data.

Read the five levels Explore the cell ecosystem

Five-level readiness ladder

A virtual cell is a sequence of testable capabilities—not one marketing label.

Observed status · no inferred scores

VC-1Available now

Evidence resource

Find and connect disease evidence without converting it into a prediction.

Input: Gene identifiers, DMD source rows, perturbation records and provenance
Output: Source-linked evidence record, uncertainty state and decision-blocking gap
Evaluation: Schema validation, checksums, source traceability and release invariants
Current evidence: 17,921 gene records · 123 detailed audit records · 5 frozen comparisons

Boundary / blockerThis layer organizes evidence; it does not estimate a DMD response.

Explore the evidence →

VC-2Partially available

Cell-state context

Place a question in a biological compartment or myogenic transition.

Input: Unperturbed human-myogenic time course and qualified context screens
Output: Reference-state placement and evidence-availability map
Evaluation: Physical timepoint QC, context coverage and explicit missing-state labels
Current evidence: 271 cells · 0/24/48/72 h · 20/21 genes measured · 9/21 in myoblast screen

Boundary / blockerNo released trajectory is both DMD-specific and perturbation-conditioned.

Open the cell ecosystem →

VC-3Available · limited

Same-context perturbation baseline

Test whether a model can recover held-out responses inside one processed assay.

Input: Observed HepG2 CRISPRi response substrate for 21 candidates
Output: Held-out 2,000-feature mean response vector
Evaluation: Repeated target-level folds · RMSE · raw and residual cosine · simple baselines
Current evidence: Small average-error improvement; direction unreliable; external transfer unsupported

Boundary / blockerSame-assay performance does not establish transport to muscle or DMD.

Inspect the benchmark →

VC-4Locked · no disease outcomes

DMD state-transition predictor

Predict molecular and functional response in matched DMD and control muscle states.

Input: Required: independent DMD/control perturbations with donor, state, time and function
Output: Desired: response distribution, functional effect, toxicity and calibrated uncertainty
Evaluation: Planned unseen donor, state, laboratory and disease holdouts with abstention
Current evidence: 0 registered predictions · 0 independent DMD perturbation outcomes

Boundary / blockerThe training and external truth objects required for this task do not exist.

See the data required →

VC-5Locked · not available

Patient digital twin

Estimate a patient-specific disease trajectory under an intervention.

Input: Required: longitudinal patient-linked state, intervention and outcome data
Output: Desired: patient-specific trajectory with uncertainty and prospective calibration
Evaluation: Required: prospective patient holdout, safety analysis and clinical governance
Current evidence: No patient-linked longitudinal perturbation trajectory is released

Boundary / blockerNo patient-specific prediction or treatment simulation is scientifically permitted.

Inspect the trajectory boundary →

Context × perturbation × output

The current coverage matrix exposes the missing middle.

Open machine contract →

Context	Cell state	Perturbation	Molecular output	Functional output	Current state
DMD evidence sources	Disease-associated observations	No matched perturbation	Source-linked direction and conflict	Not measured	Evidence only
Human myogenic reference	0 / 24 / 48 / 72 h	Unperturbed	Reference program trajectory	Differentiation context only	Measured reference
HepG2	One processed assay context	21 observed CRISPRi targets	2,000-feature response vector	Not a muscle endpoint	Limited baseline
DMD muscle	Myoblast → mature/function	Candidate perturbation	Required but unavailable	Required but unavailable	Not yet tested
FAP · immune · vascular	Disease microenvironment	Cell-specific perturbation	Required but unavailable	Cross-cell consequence unavailable	Not yet tested
Individual patient	Longitudinal disease state	Therapy or intervention	No calibrated output	No clinical outcome prediction	Not available

One result · three reading levels

Choose the explanation that matches your task.

Biologist

The platform can show where a gene has been observed, where the evidence disagrees and which cell experiment would resolve the uncertainty. It cannot yet tell you how DMD muscle will respond.

The next scientific unlock

Return one preregistered DMD perturbation outcome through the full evidence loop.

Study Card → registered prediction → matched experiment → positive, null, toxic or discordant outcome → governed evidence transition.

Open the decision queue