Flow2Spatial reconstructs spatial proteomics through transfer learning

Flow2Spatial is the computational part of PLATO (parallel flow projection and transfer learning across omics data).

It aims to reconstruct spatial proteomics from the values of parallel-flow projections in POTTLE. Leveraging transfer learning, Flow2Spatial can restore fine structure of protein spatial distribution in different tissue types.

_images/Flow2Spatial.png

Key functions of Flow2Spatial

  • Three data generators: generator.omics(), generator.histology() and generator.random().

  • Reconstruction model training: model.preparation(), model.training() and model.reconstruction().