Spatial simulations of cell processes require realistic cell geometries — accurate representations of where organelles are, how they’re shaped, and how they vary from cell to cell. Building those geometries by hand doesn’t scale. CellOrganizer was designed to learn them directly from microscope images.
This chapter describes the full workflow: from image preparation through model training, quality assessment, geometry sampling, and integration with biochemical simulation frameworks.
What CellOrganizer Does
CellOrganizer learns generative statistical models of cell spatial organization from fluorescence microscopy images. A trained model captures not just a single representative cell shape, but the full distribution of variation — in overall cell architecture, organelle count, organelle size and shape, and spatial positioning — across a population of imaged cells.
From that model, new cell geometries can be sampled on demand. Each sampled geometry is a statistically plausible instance of the learned cell type, ready to serve as the spatial scaffold for a biochemical simulation.
The Workflow
- Image preparation — preprocessing fluorescence images for model training
- Model learning — fitting generative models to the image data
- Quality assessment — evaluating how well the learned model represents the input data
- Geometry sampling — generating new cell instances from the trained model via multiple approaches (mesh, SBML-spatial, VCML)
- Simulation integration — pairing sampled geometries with biochemical specifications to run spatial simulations
The result is a pipeline that takes raw microscopy data as input and produces simulation-ready cell geometries as output, with statistical variation preserved throughout.
Why It Matters
Biochemical processes in cells are inherently spatial. Reaction rates, signaling gradients, and transport dynamics all depend on where molecules are relative to organelles and membrane boundaries. Simulations that ignore geometry — or use a single hand-drawn geometry — miss that variation entirely. CellOrganizer makes it practical to run spatial simulations across a realistic ensemble of cell shapes, enabling more robust and biologically meaningful results.
This work came out of the Murphy Lab at Carnegie Mellon University, where I contributed to CellOrganizer’s development as part of the computational biology group.