Building virtual-cell dynamics
that evolve with biological measurements

Virtual cells must do more than represent observed states — they must predict what happens next. CellCompass determines when existing dynamics are sufficient and when evolving measurements require a new model to be designed, tested, and retained.

Get started See what it designed
  • 100DynBench scenarios
  • 4measurement constraints
  • 4real single-cell studies
  • ∞long-horizon research
In motion

From biological measurements to designed dynamics.

A hand-drawn sketch of the idea — a cell on the bench, an atlas streaming into the machine, and the design cycle that draws new dynamics as new questions arrive.

a moving sketch · sound on for the score

Grounded in benchmark and real-data evidence

DynBench100 controlled tasks HCT116 pulse-labeled drug response LARRY clone-informed hematopoiesis CITE-seq RNA-to-protein dynamics CytoBridge dynamics engine

The premise

As measurements evolve, the model must evolve with them.

Single-cell technologies can now capture cellular state, history, response, and molecular context at scale. But the dynamical models used to predict future states generally remain fixed. This creates a growing gap between what experiments can reveal and what virtual cells can predict.

Figure 1 system overview showing biological questions, evolving measurements, the CellCompass model-design loop, and reusable dynamics families.
Figure 1 · system overview Experimental evidence reshapes the dynamics base.
01

Prediction beyond observed states

Virtual cells must forecast cellular change under biological context and intervention.

02

The predictive task is not fixed

Each new experiment can expose a relation that existing dynamics cannot express.

03

Model design must accumulate

Scientific capacity should grow across studies instead of restarting from one-off solutions.

The modeling principle

New measurements change what a faithful cellular future must preserve.

New measurements do more than enrich a cellular representation. CellCompass translates the biological relation they reveal into an explicit constraint on future dynamics — movement, population change, lineage divergence, perturbation response, or future molecular phenotype.
01
Time course

Velocity through state space

Sampled marginals constrain how populations move between observed time points.

transport
02
Clone record

Branching future fate

Lineage barcodes force the dynamics to preserve hidden fate divergence.

fate mass
03
Growth / death

Changing population mass

Proliferation and apoptosis reshape balanced transport into mass-changing dynamics.

unbalanced
04
Multi-omics

Coupled regulatory state

RNA and protein readouts require trajectories that carry molecular programs across modalities.

RNA → protein
The distinctive capability

Virtual cells have scaled representations. CellCompass scales the dynamics themselves.

CellCompass extends agentic AI from analyzing data and choosing among fixed models to constructing, testing, and retaining mechanistically structured dynamics. Pulse labeling, lineage tracing, and paired RNA–protein measurements each exposed a relation that existing dynamics could not preserve — and CellCompass converted those gaps into reusable models.

1adequate model reused
3new dynamics designed
0.702highest DynBench aggregate score
New measurement Adequacy decision Construct Test Retain for reuse
Metabolic RNA labeling 01

PulseBridge

Newly synthesized and pre-existing RNA became constraints on staged cancer drug-response dynamics.

−31.1% PCIE 0.680 vs. 0.988 · mean W1 0.301 vs. 0.308 · 5.3 h
Clone tracing 02

LineageBridge

Clone records became fate dynamics that preserve hidden divergence between transcriptionally similar cells.

+54% CFCorr 0.427 vs. 0.277 for the strongest baseline · 20.8 h
CITE-seq multi-omics 03

ProteinBridge

Paired RNA–protein measurements became RNA-initialized dynamics that generate future surface-protein states while preserving paired structure.

−13.2% ProteinEnergy 0.0222 vs. 0.0256 · covariance error −11.1% · 28.2 h
The equally important decision It does not invent a new model when the right dynamics already exist.

For epithelial-to-mesenchymal transition, CellCompass retained the existing velocity–growth formulation and reconstructed 12 context-specific trajectories across four cancer cell lines and three stimuli.

From analysis to model building

An AI system for cumulative dynamical-model building.

CellCompass moves beyond analyzing data, writing code, or selecting among fixed models. It constructs new mechanistically structured dynamics, evaluates them against the biological relation that motivated them, and turns each confirmed result into a reusable capability.

Autonomous algorithm design

Proposes, implements, and tunes brand-new dynamical-modeling algorithms inside isolated, version-controlled workspaces.

Theory-grounded reasoning

Literature & theory-book RAG over optimal transport and flow matching informs every modeling decision with citations.

Reusable scientific capacity

New methods register as first-class models on the open CytoBridge engine — reusable for simulation and downstream analysis across datasets.

Measurement-grounded evaluation

Every candidate is tested against the future relation revealed by the experiment, with automatic promote / reject against baseline gates.

Reproducible by design

A registry tracks every proposal, snapshot, decision, and locked release as the single source of truth.

Long-horizon autonomy

Guarded subprocesses, timeout & OOM protection, and durable checkpoints let a campaign run for days and recover itself — the agent never crashes.

The dynamics framework

Four forces of life, one transport equation.

CytoBridge supplies the mechanistically structured foundation that CellCompass can reuse or extend. Velocity, growth, stochasticity, and interaction are explicit modeling choices rather than hidden behavior inside a black-box predictor.

RUOT continuity equation learned from snapshots
∂tρt + ∇·(ρt vt) + ∇·(ρt Kρt) = ½σ²Δρt + gt ρt
velocity field v growth rate g score / diffusion σ interaction kernel Kρ

The learned drift, growth, diffusion, and density-dependent interaction kernel transport ρt between observed time-points. The interaction term follows the mean-field McKean–Vlasov formulation for collective cell dynamics.

Model family
Dynamical OT Unbalanced OT RUOT CRUFM UMFSB CytoBridge
Spatiotemporal dynamical generative model: cells differentiate, grow, die, and interact across time, modeled as transport between an initial and target distribution.

The spatiotemporal dynamical generative model — velocity, growth, score, and interaction transport an initial distribution to later time-points.

Measurement-grounded evaluation

Tests whether an agent can build dynamics, not just write code.

DynBench100 uses controlled ground truth to test dynamical reasoning across velocity, growth, held-out distributions, fate probabilities, perturbation response, and gene-regulatory recovery. CellCompass achieved the highest aggregate performance among the tested coding and biomedical agents.

100scenarios across easy, medium, hard, and adversarial settings
0.702highest aggregate score among tested agents
6dynamical objectives evaluated per agent
20 × 3selected scenarios × repeats; *** adjusted P < 0.001
DynBench evaluates CellCompass across controlled dynamical modeling tasks, showing task design, difficulty levels, baseline ranking, task profiles, difficulty scaling, BRGU-SFM component performance, and repeat stability.
DynBench100 probes six biological objectives and increasing difficulty; CellCompass leads baseline agents and selects BRGU-SFM as a stronger overall trade-off.
Cumulative model building

Construct. Test. Retain. Repeat.

A new measurement exposes a predictive gap; CellCompass proposes and implements a formulation, tests it against the motivating relation, and locks the validated model into the reusable dynamics base.

01 · 07 Motivation

A real biological gap

A dataset where growth, stochasticity, or cell–cell interaction is poorly captured by existing dynamical models.

    Reviewer-gated. A proposal clears review before any code is written; baseline gates then promote or reject every stage automatically.
    Long-horizon. Isolated execution, durable checkpoints, and self-recovery let a campaign run for days — continuously debugging itself.
    Registry-first. Stages, gates, archived trials, and locked releases live in a registry — the source of truth, not the chat.
    Honest. Metrics use model-generated trajectories from t₀, never future observed targets — no metric hacks.
    Campaign replay

    From a missing biological relation to a locked model.

    Each replay follows a design campaign from the motivating experimental relation through competing formulations, scored trials, and baseline gates to a confirmed release. Every trial is retained, making the path to PulseBridge, LineageBridge, and ProteinBridge inspectable rather than a single-prompt result.

    GSE305370 · CITE-seq time course

    ProteinBridge turned RNA-only dynamics into a protein-state generator.

    CellCompass designed ProteinBridge around CITE-PCFM: a model that starts from RNA and generates time-resolved ADT surface-protein states.

    What changed during search
      Biological problem Input question and measurement constraints
      CITE-seq panel showing the biological request and CellCompass RNA-to-protein formulation.
      Output analysis Downstream biological result
      CITE-seq panel showing CellCompass-designed RNA-to-protein dynamics and downstream protein-fate programs.
      Optimization history

      Surface-protein distribution error

      trial 1 / 70 0.3233
      high low real-data trial order
      trial result best retained so far lower is better
      Cross-design search landscape

      CellCompass moved between algorithm basins before locking CITE-PCFM.

      The upper chart zooms into CITE-PCFM tuning. This panel replays the broader search across candidate algorithm families.

      Model reuse, not just invention

      Sometimes the right model already exists.

      The adequacy decision is as important as invention. For EMT across four cancer cell lines and three stimuli, CellCompass judged the existing VGFM velocity–growth formulation sufficient, then focused its effort on fitting, evaluation, and biological interpretation.

      • 12 trajectoriesseparate velocity–growth dynamics for every cell-line × stimulus pair — 34,994 cells, no new algorithm required.
      • Context-gated routesthe same stimulus yields mesenchymal conversion, partial/hybrid EMT, or epithelial retention depending on the initial cell-line state.
      • Three-layer mechanismstimulus-proximal pathways, cell-state gating, and growth/mass coupling explain why the routes separate.
      Generated endpoint route atlas across twelve cell-line and stimulus conditions, showing mesenchymal conversion, partial or hybrid EMT, epithelial-retaining and mixed routes.
      Generated endpoint EMT-core shift vs. partial-EMT/hybrid shift across all twelve cell-line × stimulus conditions.
      How scientific capacity grows

      From a measured relation to a retained capability.

      Formulate the future

      Define the biological question and identify the relation between present and future cells.

      Inspect the measurement

      Verify the data and determine which temporal, lineage, perturbation, or multi-omic structure is informative.

      Reuse or design

      Reuse an adequate formulation or construct the mathematical structure that the new relation requires.

      Test the claim

      Evaluate generated futures against the motivating measurement and the strongest available baselines.

      Retain the model

      Lock the confirmed formulation into CytoBridge as a versioned capability available for future studies.

      Report the biology

      Turn model-generated trajectories into mechanisms, perturbation hypotheses, figures, and a reproducible report.

      Inspect every scientific campaign

      One model-building system, three working surfaces.

      Terminal, TUI, and Web UI expose the same versioned campaign: the biological motivation, proposals, archived trials, evaluation gates, and the final retained formulation.

      • LangGraph checkpointing & event-based timeline restore
      • Provider-aware auth: API key, Gemini & Codex OAuth
      • Boundary-based context compaction for long runs
      # Reconstruct dynamics, then report — in one turn
$ cellcompass run data.h5ad \
    --question "Analyze the main fate branches and driver genes" \
    --output ./run_demo \
    --device cuda

✓ preprocessing · AnnData contracts verified
✓ theory selected · Regularized Unbalanced OT
✓ training · velocity + growth + score
✓ downstream · fate branches, driver genes
✓ report · ./run_demo/report.html
      # Converse with the agent, resume any time
$ cellcompass

CellCompass › Check the dataset first, then propose a workflow.
  ↳ /status   /model   /compact   /resume   /jobs

› Design a new growth-aware flow-matching algorithm
✓ proposal drafted · review & campaign started
✓ trial 3/8 promoted · beats baseline gate
      # Full Web UI with live timeline replay
$ cellcompass web --host 0.0.0.0 --port 8000

✓ runtime ready · FastAPI lifespan hooks installed
→ open http://localhost:8000

# timeline · proposal review · skill toggles · compaction
      ∞long-horizon campaigns — resumable, and self-recovering for days
      1registry of record for every proposal, gate & locked release
      0metric hacks — inference always starts from t₀
      The research behind it

      Built on a line of published theory.

      npj Syst. Biol. Appl. · 2025 Deciphering cell-fate trajectories using spatiotemporal single-cell transcriptomic data
      ICLR · 2025 · Oral Learning stochastic dynamics from snapshots through regularized unbalanced optimal transport
      NeurIPS · 2025 Modeling cell dynamics and interactions with the Unbalanced Mean-Field Schrödinger Bridge
      NeurIPS · 2025 Joint Velocity-Growth Flow Matching for single-cell dynamics modeling
      NeurIPS · 2025 Variational Regularized Unbalanced Optimal Transport: single network, least action
      Nature Methods stVCR: reconstructing spatio-temporal dynamics of cell development using optimal transport
      Core team

      The core team behind CellCompass.

      The initial CellCompass team brings together single-cell dynamics, mathematical modeling, and AI-assisted model design.

      Peking University
      University of California, Irvine
      Qing Nie

      Qing Nie

      University of California, Irvine

      Bring a new measurement.
      Let the dynamics evolve.

      Reuse an existing model when it is adequate, or begin a measurement-grounded design campaign.

      install
      # 1 · install the agent + core engine
$ pip install -e . && pip install -r requirements-agent.txt
$ cd CytoBridge-main && pip install -e .

# 2 · start exploring
$ cellcompass "Analyze the available dataset and propose the next workflow"
      View CytoBridge on GitHub Read the docs