miniMTI: minimal multiplex tissue imaging enhances biomarker expression prediction from histology

Area of Science:

• Computational pathology
• Biomedical imaging
• Molecular pathology

Background:

• Virtual multiplexing using histology (H&E) is advancing but limited in molecular insight.
• Morphology-based inference alone has inherent limitations for comprehensive molecular state determination.

Purpose of the Study:

• To introduce miniMTI, a framework for molecularly anchored virtual staining.
• To identify the minimal set of molecular markers needed to reconstruct multiplex tissue imaging (MTI) panels from H&E images.
• To preserve biological and clinical information during virtual staining.

Main Methods:

• Utilized paired same-section H&E and MTI data.
• Employed a unified multimodal generative model.
• Incorporated an iterative panel selection strategy to identify informative molecular anchors.

Main Results:

• miniMTI reduced a 40-marker MTI assay to H&E plus as few as three molecular markers across colorectal and prostate cancer cohorts.
• Accurately recovered withheld markers, preserved cellular phenotypes, spatial architecture, and disease-associated molecular programs.
• Demonstrated effectiveness across two MTI platforms and over 40 million cells.

Conclusions:

• miniMTI integrates histology with sparse molecular data, overcoming limitations of morphology-only virtual staining.
• Provides a scalable and cost-effective method to expand MTI biomarker coverage.
• Retains biological interpretability and clinical relevance for virtual staining applications.