Generalizable virtual staining from histopathology to spatial proteomics across biological scales

QUEST enables virtual staining that preserves spatial protein marker expression from pixels to patients, trained on a heterogeneous atlas of paired H&E and MIF images with over 10,000 tissue regions.

• Why it matters: Routine H&E archives contain morphological correlates of molecular states but lack direct spatial protein measurements, limiting how spatial proteomics can be leveraged at scale.
• What they did: The framework represents protein markers as semantic queries for flexible multiplex immunofluorescence prediction, learning from paired H&E–MIF data across diverse tissue types and cohorts.
• The result: Virtual staining retained biologically meaningful information for cell phenotyping, neighborhood analysis, tissue organization, and patient stratification, and H&E-based queries retrieved MIF references with similar molecular profiles.

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