MIAMI

MIAMI quantifies co-expression among imaging markers using mutual information to assess statistical dependencies of marker intensities in multiplex imaging studies.

Key Features:

• Mutual Information Metric: Utilizes mutual information (MI) as a metric of co-expression between marker intensities, assessing statistical dependency and avoiding manual thresholding.
• Efficient Estimation Technique: Employs an alternative formulation and a new generalization of MI that enables efficient estimation without explicit joint density computation, improving scalability to multiple markers.
• Robustness through Simulation Studies: Validated robustness across different scenarios using simulation studies.

Scientific Applications:

• Lung Cancer Research: Applied to a lung cancer dataset to identify significant co-expression between HLA-DR and CK that is associated with patient survival.
• Triple Negative Breast Cancer Analysis: Identified co-expression among immuno-regulatory proteins PD1, PD-L1, Lag3, and IDO linked to disease recurrence in triple-negative breast cancer.

Methodology:

Redefines mutual information via an alternative generalization and estimation formulation to avoid joint density estimation and emphasize dependency structure between markers rather than absolute intensities.