MitoDecode: quantitative phenotypic decoding of mitochondrial stress states from fluorescence images

MitoDecode deep learning quantifies mitochondrial stress type and severity from fluorescence microscopy images, producing per-cell phenotypic probability profiles without segmentation across diffraction-limited and super-resolution modalities.

• Why it matters: Measuring mitochondrial stress from fluorescence images is difficult because prior approaches depend on segmentation, hand-crafted features, and discrete severity classes that can introduce error and bias.
• What they did: The method uses a parameter-free network to infer stress phenotypes directly from images, generating continuous probability profiles across treatment doses and working across multiple cell types, imaging modalities, and stress conditions.
• The result: Validated with biochemical assays, MitoDecode reliably detects mitochondrial phenotypes and enables phenotypic stress profiling of drug treatments, including cases where stressors target non-mitochondrial pathways.

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