Siyu He

Department of Biomedical Data Science at Stanford University, USA

Cell identity and fate are governed by the dynamic information flow encoded  in  cellular  transcriptomes  across  time  and  space.  While current high-throughput molecular profiling enables characterization of cell  identity,  limitations  in  spatiotemporal  resolution  restrict  the ability to model various dynamic processes. Advances in generative AI combined  with  massive  single-cell  datasets  across  species  and engineered tissues make it possible to build foundation models for

virtual cells. Here we present Squidi , a diffusion model based generative framework that predicts transcriptomic changes across diverse cell types in  response to environmental changes. We demonstrate the robustness of Squidi across cell differentiation, gene perturbation and drug response  prediction. Through continuous denoising and semantic feature integration, Squidi learns transient cell states and predicts high-resolution transcriptomic landscapes over time and conditions. Furthermore, we applied Squidi to model blood vessel organoid development and cellular responses to neutron irradiation and growth factors. Our results demonstrate that Squidi enables in silico screening of molecular landscapes and cellular state transitions, facilitating rapid hypothesis generation and providing valuable insights into the regulatory principles of cell fate decisions.