Fig. 1

Potential applications in oncology. a Pattern diagram for building AIVCs using biological observation data. b In silico experimentation establishing new cancer research paradigms. By inputting specific tumor observational data, multi-scale tumor simulation can be achieved computationally. Integration with VIs enables virtual experiments and guides wet-lab experimentation, with feedback promoting continuous AIVCs iteration. c Enhancement of current tumor multi-omics analysis. By learning biological characteristics across numerous time points during tumor cell evolution, interpolation between discrete time points enables the reconstruction of tumor cell dynamic changes. Integration of generative AI methods into AIVCs expands scarce biological sample data. d Clinical applications. Standard digital patient models for specific diseases are constructed using extensive clinical diagnostic and treatment data. Individual patient models are created through novel low-cost diagnostic data, enabling virtual treatment simulations to predict efficacy and adverse reactions, better guiding actual drug administration. AIVCs artificial intelligence virtual cells, T1/2/3 actual process time points, pT1/2 AIVC-simulated pseudotime, URs universal representations, VIs virtual instruments