Latent Dynamics · 2026-10-06
VCLMU: Mechanism-Centric Virtual Cell World Modeling for Perturbation Response
Yuwei Miao, Azim Dehghani Amirabad, Scott Oloff, Junzhou Huang, Tianyu Cui, Rui Liao
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
VCLMU is a mechanism-centric virtual cell world model that predicts perturbation response via action-conditioned stochastic latent transitions. Cellular state is represented as a sparse set of Latent Mechanism Units (LMUs) with reusable, evidence-grounded identities and observation-specific states; perturbations act as actions that induce mechanism-specific transitions before decoding transcriptional responses. Trained with two-stage pretraining (pseudo-bulk then gene-aligned single-cell), it improves perturbation-specific recovery across six perturbation-disjoint benchmarks.
Why it matters
The authors argue that many perturbation predictors are observation-centric and leave the intervention-induced latent state transition implicit. VCLMU formulates perturbation response as action-conditioned latent-state transitions over mechanism-factorized units, aiming to improve response recovery and yield interpretable, mechanism-level representations aligned with biological response programs.
Method
- Represent each cell as a sparse set of LMUs: a bank of K LMUs with shared evidence-grounded identities (candidate gene sets) and observation-specific states, routed using only the control profile.
- Encode each genetic perturbation once as an action embedding from the target gene and GRN neighborhood, then apply it independently to each active LMU to induce residual stochastic transitions.
- Train with two-stage pretraining (pseudo-bulk then gene-aligned single-cell) and supervise latent dynamics via stop-gradient target alignment and observation grounding, while decoding the transitioned states into a response distribution.
Limitation
The current study focuses primarily on single-gene transcriptomic perturbations, and future work will extend LMU validation to multimodal and combinatorial perturbation settings.
Abstract (from arXiv)
Predicting cellular responses to genetic perturbations is a central capability for virtual cells and a key step toward computational modeling of biological interventions. Most existing models directly map an unperturbed molecular profile and perturba- tion to the resulting observation without explicitly representing the latent cellular transition induced by the intervention. We introduce a mechanism-centric virtual cell world model that represents cellular state as a set of Latent Mechanism Units (LMUs) and treats genetic perturbations as actions on these latent states. Each LMU combines a reusable identity grounded in multimodal biological evidence with an observation-specific state, allowing a perturbation to induce mechanism- specific stochastic transitions before decoding the resulting transcriptional response. We train VCLMU through two-stage pretraining, first on around 200K pseudo-bulk perturbation profiles and then on gene-aligned single-cell perturbation data. Across six perturbation-disjoint benchmarks, VCLMU consistently improves perturbation- specific response recovery over strong baselines while maintaining competitive global response accuracy. We further analyze learned LMUs through enrichment between perturbation responses and LMU gene sets and show that they capture structured biological response programs. These results support mechanism-level latent state transition as a useful formulation for virtual cell models that aim to predict and interpret cellular responses to biological interventions.
Related papers
- The Delta Gap: Cellular World Models for Perturbation Response Prediction Are Far Harder Than Global Correlation Implies
- CellCanon: An Agentic Cellular World Model for Prediction, Explanation, and Observation-Driven Updating
- P2P: Cross-View Population Denoising for Unpaired Single-Cell Perturbation Response Prediction
- NeuronDiscover: Agent-in-Twin for Mechanistic Discovery in Neuronal Microenvironments with World Action Models
- KoopCell: Koopman-Based Generative Model for Learning Single-Cell Dynamics from Distribution Snapshots
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-03 · Indexed by Pub-AI: 2026-10-06 UTC
VCLMU: Mechanism-Centric Virtual Cell World Modeling for Perturbation Response
Preprint posted: 2026-10-03 · Original record ↗
- arxiv
- 2610.04475
Explicit source versions (1)
- v1 · 2026-10-03
- Preprint available
Source date: 2026-10-03 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Predicting cellular responses to genetic perturbations is a central capability for virtual cells and a key step toward computational modeling of biological interventions. Most existing models directly map an unperturbed molecular profile and perturba- tion to the resulting observation without explicitly representing the latent cellular transition induced by the intervention. We introduce a mechanism-centric virtual cell world model that represents cellular state as a set of Latent Mechanism Units (LMUs) and treats genetic perturbations as actions on these latent states. Each LMU combines a reusable identity grounded in multimodal biological evidence with an observation-specific state, allowing a perturbation to induce mechanism- specific stochastic transitions before decoding the resulting transcriptional response. We train VCLMU through two-stage pretraining, first on around 200K pseudo-bulk perturbation profiles and then on gene-aligned single-cell perturbation data. Across six perturbation-disjoint benchmarks, VCLMU consistently improves perturbation- specific response recovery over strong baselines while maintaining competitive global response accuracy. We further analyze learned LMUs through enrichment between perturbation responses and LMU gene sets and show that they capture structured biological response programs. These results support mechanism-level latent state transition as a useful formulation for virtual cell models that aim to predict and interpret cellular responses to biological interventions.
