MIAAIM: Multi-omics image integration with dimensional reduction for tissue state mapping

Area of Science:

• Biomedical Imaging
• Computational Biology
• Molecular Pathology

Background:

• High-parameter tissue imaging offers detailed single-cell molecular analysis within spatial contexts.
• Integrating data from diverse imaging platforms with varying resolutions and biomolecule quantification remains a challenge for comprehensive spatial profiling.

Purpose of the Study:

• To introduce MIAAIM (Multi-omics Image Alignment and Analysis by Information Manifolds), a modular framework for aligning and processing data from distinct imaging technologies.
• To enable unbiased molecular phenotype identification correlated with cell identities and states using high-resolution immunodetection.
• To construct multimodal tissue imaging datasets for analyzing complex tissue states and cell-molecule associations.

Main Methods:

• Development of MIAAIM, a modular framework for aligning and processing multi-modal imaging data.
• Application to clinical biopsies including histological staining, imaging mass cytometry, and mass spectrometry imaging.
• Integration of spatial features from segmented cells across modalities for analysis.

Main Results:

• MIAAIM successfully aligned and analyzed data from disparate imaging platforms with varying resolutions and complexities.
• In diabetic foot ulcer biopsies, unique molecular characteristics of immune cells were identified based on local tissue health.
• In prostate cancer TMAs, MIAAIM achieved >90% accuracy in classifying tumor grades, largely driven by spatial features, and identified novel immune signatures.

Conclusions:

• MIAAIM provides a general, agnostic framework for constructing multimodal tissue imaging datasets.
• The framework facilitates novel insights into molecular analyte associations with cell subsets and activation states.
• MIAAIM enhances the analysis of complex tissue states and disease mechanisms through integrated spatial and molecular profiling.