The paper introduces Squidiff, a novel computational framework based on a conditional diffusion model with a semantic encoder, designed to predict single-cell transcriptomic changes in response to various environmental stimuli. This predictive model addresses the limitations of previous methods by learning complex, high-resolution dynamic cellular states across different scenarios, including cell differentiation, gene perturbations, and drug responses. The authors demonstrate Squidiff’s efficacy by applying it to blood vessel organoid (BVO) development, successfully modeling the effects of both neutron irradiation and the radioprotective drug G-CSF. Overall, Squidiff provides an in-silico tool for screening molecular landscapes and inferring transient cell state transitions, aiding in the discovery of regulatory principles governing cell fate.

References:

• He S, Zhu Y, Tavakol D N, et al. Squidiff: Predicting cellular development and responses to perturbations using a diffusion model[J]. Nature Methods, 2025: 1-13.