VCLMU: Mechanism-Centric Virtual Cell World Modeling for Perturbation Response

2026-10-03

AI Summary

This study addresses the limited predictive and interpretive capacity of existing models, which lack explicit representations of cellular state transition mechanisms under genetic perturbations. We propose a mechanism-centric virtual cell world model that represents cellular states as Latent Mechanistic Units (LMUs) and treats genetic perturbations as actions operating on these units, thereby simulating mechanism-level stochastic state transitions. By integrating a reusable identity module with observation-specific state architectures grounded in multimodal evidence, our approach transcends conventional direct-mapping paradigms. Training follows a two-stage strategy comprising large-scale pseudo-bulk perturbation profile pretraining followed by single-cell data fine-tuning. Evaluated across six disjoint benchmarks, the proposed model significantly improves the accuracy of perturbation-specific response recovery, demonstrating the capacity of LMUs to capture structured biological programs.

Abstract

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.

Problem

• virtual cell
• genetic perturbation
• perturbation response
• cellular state transition

Innovation

• Virtual Cell World Model
• Latent Mechanism Units
• Genetic Perturbation
• Stochastic State Transition
• Two-stage Pretraining