Chreode: A Cell World Model for One-Step Temporal Dynamics and Perturbation Prediction

2026-05-27

AI Summary

Accurately predicting transcriptional state changes in cells under developmental cues or genetic perturbations remains a central challenge for in silico cell simulation. This work proposes Chreode, a single-step cellular world model that enables action-conditioned state transitions via a structured residual transition operator, shifting distributional evolution to the training phase to support one-step forward generation. Innovatively decomposing the Waddington landscape into downhill flow, tangential rotation, and stochastic diffusion components, Chreode integrates a shared scVI encoder with a DiT-based dynamical backbone and is pretrained on a mouse embryonic atlas comprising 2.4 million cells, enabling cross-task transfer. The model outperforms baselines on the Weinreb and Veres datasets, reduces DE20 error by 12.4% on Norman Perturb-seq when used as an embedding for GEARS, and achieves zero-shot clonal fate prediction performance comparable to dynamic optimal transport methods.

Abstract

Predicting how a cell will change its transcriptional state under a developmental signal or a genetic perturbation is the computational core of in-silico biology and the AI Virtual Cell program. Existing approaches either fit static control-to-treated maps that discard time, or solve multi-step ODE / Schrödinger-bridge problems on each dataset independently. We introduce Chreode, a one-step cell world model that predicts action-conditioned cell-state transitions through a structured residual transition operator. It shifts distributional evolution from inference time to training time, enabling single-pass generation while preserving a Waddington-inspired decomposition into downhill landscape flow, rotational in-tangent dynamics, and stochastic spread. The model is pretrained with a shared scVI encoder and a DiT-based dynamics backbone on a 2.4M-cell mouse embryonic atlas spanning 7 datasets. As a fine-tuning initialization, Chreode improves per-target Sinkhorn distance on Weinreb hematopoiesis and Veres islet differentiation over matched scratch models, PI-SDE, and PRESCIENT. As a transferable gene-state embedding for GEARS, the pretrained dynamics representation reduces shared-vocabulary DE20 mean squared error on Norman Perturb-seq from 0.2121 to 0.1858, a 12.4% relative improvement, without changing the GEARS training procedure. We interpret this transfer to perturbation prediction as evidence that pretrained developmental-trajectory dynamics encode differentiation primitives transferable to CRISPR-induced state shifts, since both involve cell-state transitions in a shared latent geometry. The pretrained backbone additionally produces zero-shot clonal fate scores on Weinreb that are competitive with strong dynamic-OT baselines.

Problem

• cell-state transition
• temporal dynamics
• genetic perturbation
• developmental trajectory
• in-silico biology

Innovation

• cell world model
• one-step dynamics
• structured residual transition
• pretrained developmental dynamics
• transferable perturbation prediction