MultiMatch: Geometry-Informed Colocalization in Multi-Color Super-Resolution Microscopy

With recent advances in multi-color super-resolution light microscopy it has become possible to simultaneously visualize multiple subunits within complex biological structures at nanometer resolution. To opti-mally evaluate and interpret spatial proximity of stainings on such an image, colocalization analysis tools have to be able to integrate prior knowledge on the local geometry of the recorded biological complex. Here, we present MultiMatch to analyze the abundance and location of chain-like particle arrangements in multi-color microscopy based on multi-marginal optimal unbalanced transport methodology. Our object-based colocalization model statistically addresses the effect of incomplete labeling efficiencies enabling inference on existent, but not fully observ-able particle chains. We showcase that MultiMatch is able to consistently recover all existing chain structures in three-color STED images of DNA origami nanorulers and outperforms established geometry-uninformed triplet colocalization methods in this task in a simulation study. Further-more, MultiMatch also excels in the evaluation of simulated four-color STED images and generalizations to even more color channels can be immediately derived from our analysis. MultiMatch is provided as a user-friendly Python package comprising intuitive colocalization visual-izations and a computationally efficient network flow implementation.