Histopathology-centered computational evolution of spatial omics: integration, mapping, and foundation models

Area of Science:

• Computational pathology
• Spatial omics
• Histopathology imaging

Background:

• Hematoxylin and eosin (H&E) imaging is the standard for morphological assessment in pathology.
• Spatial omics (SO) provides spatially resolved molecular profiling.
• Computational methods are increasingly integrating H&E images into SO analysis, aiming for single-cell resolution.

Purpose of the Study:

• To systematically review the computational evolution of spatial omics from a histopathology-centered perspective.
• To organize computational methods into distinct paradigms.
• To provide a roadmap for computational frameworks in spatial omics.

Main Methods:

• Review and categorization of computational methods in spatial omics.
• Organized methods into three paradigms: integration, mapping, and foundation models.
• Analysis focused on the integration of H&E imaging with molecular data.

Main Results:

• Identified three main computational paradigms: integration, mapping, and foundation models.
• Integration involves jointly modeling paired multimodal data.
• Mapping infers molecular profiles from H&E images, while foundation models learn generalizable representations.

Conclusions:

• Computational frameworks are crucial for advancing spatial omics in histopathology.
• Actionable modeling directions and persistent gaps were identified.
• The review provides a roadmap for developing and applying computational frameworks in spatial omics.