Kymograph classifiers predict CMT2A disease-relevant mitochondrial motility phenotypes in iPSC motor neurons

A vision transformer classifier using mitochondrial kymographs more accurately distinguishes healthy MFN2 wild-type iPSC motor neurons from MFN2 R364W-mutant diseased samples than kymorph-derived summary statistics.

• Why it matters: Scalable, robust computation is needed to quantify mitochondrial trafficking phenotypes in iPSC motor neuron models for in vitro CMT2A therapeutic screening.
• What they did: The authors trained a ViT-based framework on kymographs of mitochondrial movement along axons, comparing its performance to directly extracted measures like mitochondrial speed and the fraction of stationary mitochondria.
• The result: The model preserves high accuracy on a biological replicate holdout set and links R364W diseased kymographs to highly variable mitochondrial sizes and frequent intersection events.

AI summary of the abstract · may contain errors · not medical advice · read the paper.