Simulation-templated photorealistic prediction of bacterial patterns

Pattern formation underlies organization and function across biological systems. A major challenge in understanding and controlling these systems lies in making accurate, photorealistic predictions of emergent patterns that reflect both the governing biophysical rules and the fine-scaled visual features observed experimentally. Such predictions are essential for interpreting experimental observations and for guiding rational design in contexts where patterning governs system behavior, such as microbial consortia, tissue morphogenesis, and spatially structured biomanufacturing. Mechanistic models can reproduce global morphological patterns but lack visual details; purely data-driven generative models produce visually plausible images that are not grounded in underlying biology. Here, we introduce a simulation-templated hybrid framework that integrates a coarse-grained, mechanistic model of Pseudomonas aeruginosa colony expansion with components of a foundation image model. Using this approach, we achieve, for the first time, computationally efficient and experimentally realistic predictions of self-organized bacterial branching patterns, including fine details such as colony texture, color gradients, and replicate level variability. By combining mechanistic simulations with the expressive power of a foundation image model, our strategy offers a generalizable path for simulation-templated photorealistic prediction of biological patterns, with applications across synthetic biology, tissue morphogenesis, and other spatially structured biological and physical systems.