Virtual Cell Modeling

Introduction

A "virtual cell" is a computational model that predicts a cell's state and its response to interventions — genetic perturbations, drugs, or environmental changes — without running the wet-lab experiment. The field spans two historically separate lineages that are now converging: mechanistic simulation (ODE/SBML kinetics, flux-balance metabolism, agent-based tissue models, and the whole-cell simulations pioneered by Karr et al.), and data-driven machine learning (single-cell foundation models pretrained on tens to hundreds of millions of cells, and perturbation-response predictors). A 2024 Cell roadmap by Bunne, Roohani, Leskovec, Quake, Regev and 30+ others crystallized this into the "AI Virtual Cell" (AIVC) vision. Independent benchmarks also identify important limits: multiple 2025 benchmarks (Kedzierska et al.; Ahlmann-Eltze et al.) show that today's large foundation models, used zero-shot or for perturbation prediction, do not yet reliably beat simple linear or mean baselines — and the Arc Institute's Virtual Cell Challenge was created in part to support controlled, held-out evaluation. The following sections compare these approaches, their applications and their limitations.

Single-Cell Foundation Models (SCFMs)

Transformer models pretrained self-supervised on massive unlabeled single-cell transcriptomes, producing a general-purpose "cell embedding" that can be fine-tuned or used zero-shot for downstream tasks (cell-type annotation, batch integration, perturbation, gene-network inference).

Model architecture

A cell (~20,000 gene counts) is tokenized and fed to a transformer with a masked-prediction pretraining objective, but the tokenization differs by model:

• scGPT (Cui et al. 2024) — value-binning tokens; a generative masked-attention objective; ~33M cells. The reference "GPT-style" SCFM. Nat Methods 2024;21(8):1470–1480. DOI 10.1038/s41592-024-02201-0.
• Geneformer (Theodoris et al. 2023) — rank-value encoding (genes ordered by expression rank, no absolute values); pretrained on ~30M cells ("Genecorpus-30M"). Application: transfer learning for network biology. Nature 2023;618:616–624. DOI 10.1038/s41586-023-06139-9.
• scFoundation / xTrimoGene (Hao et al. 2024) — continuous value embedding (no lossy binning), an asymmetric encoder–decoder processing only non-zero genes, and Read-Depth-Aware (RDA) pretraining; ~100M params, >50M human cells, 19,264 genes. Nat Methods 2024;21(8):1481–1491. DOI 10.1038/s41592-024-02305-7. (The backbone BioMap adapted to win VCC 2025.)
• scBERT (Yang et al. 2022) — early BERT-style SCFM for cell-type annotation. Nat Mach Intell 2022;4:852–866. DOI 10.1038/s42256-022-00534-z.
• UCE (Universal Cell Embeddings) (Rosen, Roohani et al. 2023) — 33-layer, ~650M-param model on >36M cells across 300+ CELLxGENE datasets, using ESM2 protein embeddings so genes from any species map into one universal space. bioRxiv 10.1101/2023.11.28.568918.
• TranscriptFormer (CZI, 2025) — cross-species generative SCFM (~112M cells, 12 species); Apache-2.0 licensed. bioRxiv 10.1101/2025.04.25.650731 (bibliographic identifier requires confirmation).

Evidence / strengths. Strong when fine-tuned on a labeled downstream task; competitive cell-type annotation, batch integration, useful embeddings; enable transfer to data-scarce settings. UCE/TranscriptFormer add cross-species generalization.

Limitations. In zero-shot use, Kedzierska et al. 2025 (Genome Biology 26:101; DOI 10.1186/s13059-025-03574-x; PMID 40251685) showed scGPT and Geneformer perform inconsistently and often worse than plain highly-variable-gene selection, scVI, or Harmony for cell-type clustering across five datasets. Recurring lesson: single-cell data is dominated by technical noise (dropout, batch), the biological signal is near-linear, and huge models overfit that noise — so curation + statistical structure often beats raw scale.

Perturbation-Response Prediction

Models that predict the transcriptome-wide response to a specific perturbation (silence gene X, apply drug Y at dose Z), including generalization to unseen single or combinatorial perturbations.

How the method works

• scGen (Lotfollahi et al. 2019) — VAE doing latent-space vector arithmetic (learn a "perturbation direction," add to unseen contexts). Nat Methods 2019;16:715–721. DOI 10.1038/s41592-019-0494-8.
• CPA (Compositional Perturbation Autoencoder) (Lotfollahi et al. 2023) — adversarially disentangles basal state, perturbation, dose, covariates into composable latent factors → drug-combination and dose-response prediction. Mol Syst Biol 2023;19:e11517. DOI 10.15252/msb.202211517.
• chemCPA (Hetzel et al. 2022) — extends CPA to unseen drugs via molecular-structure encoding. NeurIPS 2022; arXiv:2204.13545.
• GEARS (Roohani et al.) — GNN over a GO / co-expression knowledge graph, predicting effects of multigene (combinatorial) perturbations including genes never perturbed alone. Nat Biotechnol 2024;42:927–935. DOI 10.1038/s41587-023-01905-6.
• STATE (Arc Institute; Adduri et al. 2025) — set-attention state-transition model on >100M perturbed cells across 70 contexts + a cell-embedding model on 167M cells; reports >50% improvement in perturbation-effect discrimination. bioRxiv 10.1101/2025.06.26.661135 (non-commercial license).
• BioMap xTrimoSCPerturb — the VCC 2025 grand-prize winner; fused the scFoundation backbone with explicit statistical baseline features and a metric-aligned (PDS+DES+MAE) loss. Documented via the Arc VCC 2025 wrap-up + BioMap's post; implementation details are developer-reported rather than independently established in this review.

Key datasets & benchmark. Genome-scale Perturb-seq — Replogle et al. 2022, Cell 185:2559–2575.e28, DOI 10.1016/j.cell.2022.05.013 — is the foundational training/transfer resource (K562/RPE1, ~8K genes). Norman et al. 2019 (combinatorial CRISPRa, GSE133344) is the standard combinatorial testbed. The Arc Virtual Cell Challenge (2025; H1-hESC, 300 CRISPRi perturbations split 150/50/100) is the flagship held-out evaluation, scored on PDS (perturbation-discrimination), DES (differential-expression), and MAE.

Strengths. GEARS and STATE genuinely push combinatorial and cross-context generalization.

Limitations. See the benchmark evaluation section — for unseen and combinatorial perturbations, deep predictors have not yet beaten additive / mean baselines in controlled benchmarks. The winning VCC recipes were hybrids (statistics + curation + selective DL), not pure end-to-end nets.

The "AI Virtual Cell" (AIVC) Vision

A field-defining roadmap: Bunne, Roohani, Rosen, … Regev, Lundberg, Leskovec, Quake et al. 2024, "How to build the virtual cell with artificial intelligence: Priorities and opportunities," Cell 2024;187:7045–7063, DOI 10.1016/j.cell.2024.11.015.

How it frames the problem. Proposes the AIVC as a multi-scale, multi-modal system of "universal representations" of molecules, cells, and tissues, learned from large atlases, that can be queried and simulated — predict unmeasured states and responses (virtual instruments / virtual perturbations) and be validated against real experiments.

Operationalization. Arc Institute turned the vision into a measurable target with the Virtual Cell Challenge — "Virtual Cell Challenge: Toward a Turing test for the virtual cell," Cell, June 26, 2025, PII S0092-8674(25)00675-0 (publisher record). Supporting infrastructure: the Arc Virtual Cell Atlas and Tahoe-100M (100M cells, ~1,100 drugs × 50 cancer lines), CZI's CELLxGENE Census and cz-benchmarks/cell-eval harness.

Strengths. Aligns a fragmented field on shared objectives, metrics, and open data; catalyzed the development of shared evaluation standards.

Limitations. Aspirational — no current system is a true virtual cell; the gap between vision and demonstrated accuracy (see benchmark evaluation) is large.

Whole-Cell & Mechanistic Models

Bottom-up simulations grounded in molecular mechanism: ODE/SBML kinetics, flux-balance metabolism (FBA), agent-based/spatial tissue models, integrative whole-cell models.

How the method works

• Whole-cell model — Karr et al. 2012, "A Whole-Cell Computational Model Predicts Phenotype from Genotype," Cell 150:389–401, DOI 10.1016/j.cell.2012.05.044. Simulates every characterized gene function of Mycoplasma genitalium via ~28 integrated submodels, predicting phenotype from genotype. The landmark proof-of-concept.
• Flux-balance analysis (FBA) — genome-scale metabolic models (Recon3D, Human-GEM) solved by linear programming under stoichiometric/steady-state constraints; tool COBRApy. (Orth et al. 2010, Nat Biotechnol 28:245–248.)
• ODE/SBML kinetic simulation — mass-action / Michaelis-Menten networks (Tellurium, AMICI, COPASI); used for signaling and cardiac safety (CiPA O'Hara-Rudy action-potential model).
• Agent-based / spatial — PhysiCell, Chaste, CompuCell3D, Morpheus, FLAME-GPU2 for tumor microenvironment, morphogenesis, drug penetration.
• Hybrid mechanistic + ML — SBML-to-JAX, adjoint-sensitivity fitting (AMICI + pyPESTO), composable multi-scale frameworks (Vivarium).

Strengths. Interpretable, causal, extrapolate to unseen conditions in principle, and are the regulatorily-accepted end of the spectrum (PBPK, CiPA/QSP appear in real FDA/EMA submissions).

Limitations. Require extensive prior knowledge and parameterization; do not scale to whole human-cell transcriptomic response; the Karr-style approach remains confined to minimal organisms and is computationally heavy.

Benchmark design and model evaluation

Rigorous benchmark studies whose central finding is that predictive performance depends on the task, dataset and evaluation design.

• Ahlmann-Eltze, Huber & Anders 2025 — "Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines," Nature Methods 22:1657–1661, DOI 10.1038/s41592-025-02772-6, PMID 40759747. For combinatorial perturbations DL did not beat a simple additive model; for unseen genes DL did not beat predicting the training mean.
• Kedzierska, Crawford, Amini & Lu 2025 — "Zero-shot evaluation reveals limitations of single-cell foundation models," Genome Biology 26:101, DOI 10.1186/s13059-025-03574-x, PMID 40251685.
• PerturBench (Wu et al. 2024) — arXiv:2408.10609 — ~27 perturbation methods across ~29 datasets with rank-based specificity metrics; strong non-DL baselines.
• The VCC as a benchmark. The Arc wrap-up reports a trivial cell-mean baseline beat most deep models on MAE, and that "pure end-to-end neural networks have yet to outperform hybrid models" — which is why the challenge weights PDS (perturbation-specificity) as the deciding metric (a constant prediction scores chance = 0.5).

The debate. Cross-paper score comparisons are usually invalid (different tasks/data/metrics). The defensible recipe the community is converging on: your own task + your own metric + a permuted null + a simple baseline + an oracle/ceiling, all on the same held-out split.

Choosing an approach

• Mechanistic models win where they're accepted: PBPK, CiPA/QSP, and FBA are the only virtual-cell approaches with real regulatory standing today.
• For perturbation-response, hybrids win — not pure deep learning. All three VCC 2025 winners (BioMap xTrimoSCPerturb; Altos flow-matching; Team Outlier's TransPert) beat brute-force scale by combining curated data + explicit statistical priors + metric-aligned loss. TransPert (3rd overall, #1 on PDS) used no neural net at all.
• Foundation models are useful but oversold. Fine-tuned they help; zero-shot they frequently lose to HVG/scVI (Kedzierska 2025). For unseen/combinatorial perturbation they don't yet beat linear baselines (Ahlmann-Eltze 2025).
• Open research questions: no system today is a true "virtual cell." STATE and the AIVC roadmap point the direction; the near-term frontier is cross-context / cross-cell-type transfer with rigorous held-out evaluation, where statistics-first hybrids currently lead.

Applying the methods with BioMate

• Mechanistic virtual-cell infrastructure (built). Four Phase-1 workflows run on BioMate's AWS-Batch engine: CiPA cardiac-safety ODE (hERG mechanistic confirmation), ODE signaling simulation, metabolic-flux FBA, and LINCS-L1000 transcriptomic MoA — integrated with existing ADMET/PBPK/RNA-seq workflows.
• Perturbation-prediction infrastructure (staged). GEARS, CPA, scGen, scFoundation/TranscriptFormer catalogued with licenses/GPU sizing/Docker recipes; several are planned Phase-2 GPU workflows.
• VCC 2026 participation (active). BioMate registered a team for the Arc Virtual Cell Challenge (starts Aug 20, 2026), with a validated scoring harness (cell-mean / additive / oracle-NN baselines) and a statistics-first plan: pseudobulk per perturbation, test-time features (baseline expression, MSigDB-C7 pathway membership, Open Targets genetics priors), similarity-weighted cross-line transfer, and global PDS scale-calibration — mirroring the TransPert recipe, adding DL only if it beats the statistical floor.
• Cross-line transfer + PDS calibration (validated at baseline level). BioMate independently reproduced the key VCC scoring behavior (constant predictions → chance PDS 0.5; nearest-neighbor borrowing → PDS ≈ 0.81 calibrated), confirming that perturbation-specific prediction and distance calibration are the high-leverage levers.
• Research scope. BioMate does not claim a trained perturbation model beating SOTA, nor a whole-cell simulator; its differentiators are the execution engine (mechanistic + ML workflows on managed AWS Batch), a rigorous baseline/eval harness, and a statistics-first VCC strategy aligned with what actually won in 2025.

References

1. Cui H, et al. scGPT. Nat Methods 2024;21(8):1470–1480. DOI 10.1038/s41592-024-02201-0.
2. Theodoris CV, et al. Geneformer. Nature 2023;618:616–624. DOI 10.1038/s41586-023-06139-9.
3. Hao M, et al. scFoundation/xTrimoGene. Nat Methods 2024;21(8):1481–1491. DOI 10.1038/s41592-024-02305-7.
4. Yang F, et al. scBERT. Nat Mach Intell 2022;4:852–866. DOI 10.1038/s42256-022-00534-z.
5. Rosen Y, Roohani Y, et al. UCE. bioRxiv 2023. DOI 10.1101/2023.11.28.568918.
6. CZI. TranscriptFormer. bioRxiv 2025. DOI 10.1101/2025.04.25.650731. [Bibliographic details require confirmation]
7. Lotfollahi M, et al. scGen. Nat Methods 2019;16:715–721. DOI 10.1038/s41592-019-0494-8.
8. Lotfollahi M, et al. CPA. Mol Syst Biol 2023;19:e11517. DOI 10.15252/msb.202211517.
9. Hetzel L, et al. chemCPA. NeurIPS 2022. arXiv:2204.13545.
10. Roohani Y, et al. GEARS. Nat Biotechnol 2024;42:927–935. DOI 10.1038/s41587-023-01905-6.
11. Replogle JM, et al. Genome-scale Perturb-seq. Cell 2022;185:2559–2575.e28. DOI 10.1016/j.cell.2022.05.013.
12. Adduri AK, et al. State (Arc). bioRxiv 2025. DOI 10.1101/2025.06.26.661135. [Non-commercial.]
13. Bunne C, Roohani Y, … Quake SR. How to build the virtual cell with AI. Cell 2024;187:7045–7063. DOI 10.1016/j.cell.2024.11.015.
14. Arc Institute. Virtual Cell Challenge: Toward a Turing test for the virtual cell. Cell, Jun 26, 2025. https://www.cell.com/cell/fulltext/S0092-8674(25)00675-0 [publisher record cited]
15. Karr JR, et al. A Whole-Cell Computational Model Predicts Phenotype from Genotype. Cell 2012;150:389–401. DOI 10.1016/j.cell.2012.05.044.
16. Ahlmann-Eltze C, Huber W, Anders S. Deep-learning perturbation prediction does not yet outperform linear baselines. Nat Methods 2025;22:1657–1661. DOI 10.1038/s41592-025-02772-6. PMID 40759747.
17. Kedzierska KZ, et al. Zero-shot evaluation reveals limitations of single-cell foundation models. Genome Biology 2025;26:101. DOI 10.1186/s13059-025-03574-x. PMID 40251685.
18. Wu Y, et al. PerturBench. arXiv:2408.10609 (2024).
19. Lopez R, et al. scVI. Nat Methods 2018;15:1053–1058. DOI 10.1038/s41592-018-0229-2.
20. Lotfollahi M, et al. scArches. Nat Biotechnol 2022;40:121–130. DOI 10.1038/s41587-021-01001-7.

Supporting resources: Arc VCC 2025 wrap-up (arcinstitute.org); Tahoe-100M / Virtual Cell Atlas; CELLxGENE Census; scPerturb (Peidli et al. 2024) Zenodo 10.5281/zenodo.13350497.