IMMUNIA: A Reasoning-Centered Framework for Immunoregulatory Surfaceome Discovery

Area of Science:

• Computational Biology
• Immunology
• Artificial Intelligence

Background:

• Current immunotherapy biomarker discovery relies heavily on correlation, often missing complex tumor immune interactions.
• Mechanistic and context-dependent aspects of tumor immunity are crucial for effective biomarker identification.

Purpose of the Study:

• To introduce IMMUNIA, a reasoning-centered AI framework for prioritizing immunoregulatory surfaceome genes.
• To enable structured, interpretable, and multi-model inference for biomarker discovery in cancer immunotherapy.

Main Methods:

• IMMUNIA integrates standardized prompting, literature context, and consensus reasoning across multiple large language models (LLMs).
• The framework evaluates candidate genes based on immunotherapy relevance, inflammation, and NF-κB signaling.
• Applied to prostate cancer transcriptomic data, analyzing 458 immunoglobulin domain-containing surfaceome genes.

Main Results:

• Systematic analysis identified a prioritized set of candidate genes through cross-model evaluation.
• Internal validation confirmed robust discrimination of immune-relevant targets.
• Consensus prioritization recovered known immunoregulatory molecules and highlighted PTPRS, VCAN, and MXRA5 for stromal-mediated immune regulation.

Conclusions:

• IMMUNIA provides a transparent and reproducible method for mechanistically informed biomarker prioritization.
• Reasoning-centered AI complements data-driven approaches for discovering immunoregulatory targets in the tumor microenvironment.
• The framework facilitates the generation of testable hypotheses by integrating existing biological knowledge.