Predicting subcellular spatial metabolomics from microscopy

ABSTRACT: Current spatial metabolomics techniques have transformed our understanding of cellular metabolism, yet accessible methods remain limited in spatial resolution owing to an intrinsic trade-off between pixel size and analyte sensitivity. Here we present MetaLens, a method for in silico spatial metabolomics (ISSM) that disrupts this trade-off by quantitatively propagating cellular-resolution matrix-assisted laser desorption/ionization (MALDI) imaging mass spectrometry (IMS) readouts to subcellular scales through integration with high-resolution light microscopy. Trained on thousands of paired MALDI and microscopy acquisitions, MetaLens learns to predict and super-resolve molecular distributions solely from optical input, enhancing effective spatial resolution by approximately 70-fold relative to the measured MALDI pixel size, while achieving high prediction accuracy across 72 analytes spanning multiple lipid classes. We validated the super-resolution and quantitative capabilities of MetaLens against both pseudo-MALDI fluorescence ground truth and independently acquired high-resolution MALDI imaging data in hepatocytes. Applied to a lipotoxic human hepatocyte model of metabolic dysfunction-associated steatohepatitis (MASH), MetaLens delineated subcellular metabolic domains with distinct molecular compositions, revealing saturation-dependent compartmentalization of sphingomyelin species as well as cytoplasmic localization of an exogenous drug compound. Because inference requires only microscopy, MetaLens preserves specimen integrity for downstream assays, enabling accessible, label-free subcellular spatial metabolomic analysis from standard optical acquisitions.

ORGANISM(S): Homo sapiens (human)

SUBMITTER: Luca Rappez 

PROVIDER: S-BSST3182 | biostudies-other |

REPOSITORIES: biostudies-other