Multimodal Oncology Agent for IDH1 Mutation Prediction in Low-Grade Glioma

2025-12-05

AI Summary

This study addresses the non-invasive prediction of isocitrate dehydrogenase 1 (IDH1) mutation status in low-grade gliomas (LGGs) to support clinical stratification and personalized treatment. We propose a novel multimodal tumor agent framework that— for the first time—integrates foundation model–driven histopathological analysis (using TITAN to extract tissue-level features from whole-slide images) with external biomedical knowledge reasoning (synthesizing clinical-genomic evidence from PubMed, Google Search, and OncoKB). This design enables synergistic, cross-modal information mining. Evaluated on the TCGA-LGG cohort, our method achieves an F1 score of 0.912, significantly outperforming unimodal baselines and conventional feature-fusion approaches. Our key contribution lies in establishing an interpretable, knowledge-augmented multimodal reasoning paradigm—delivering high-accuracy, clinically deployable IDH1 status prediction with transparent decision support.

Technology Category

• Machine Learning: Multimodal Learning
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Low-grade gliomas frequently present IDH1 mutations that define clinically distinct subgroups with specific prognostic and therapeutic implications. This work introduces a Multimodal Oncology Agent (MOA) integrating a histology tool based on the TITAN foundation model for IDH1 mutation prediction in low-grade glioma, combined with reasoning over structured clinical and genomic inputs through PubMed, Google Search, and OncoKB. MOA reports were quantitatively evaluated on 488 patients from the TCGA-LGG cohort against clinical and histology baselines. MOA without the histology tool outperformed the clinical baseline, achieving an F1-score of 0.826 compared to 0.798. When fused with histology features, MOA reached the highest performance with an F1-score of 0.912, exceeding both the histology baseline at 0.894 and the fused histology-clinical baseline at 0.897. These results demonstrate that the proposed agent captures complementary mutation-relevant information enriched through external biomedical sources, enabling accurate IDH1 mutation prediction.

Problem

• Predict IDH1 mutations in low-grade glioma using multimodal data.
• Integrate histology, clinical, and genomic information for accurate prediction.
• Enhance prediction by reasoning over external biomedical knowledge sources.

Innovation

• Integrates TITAN foundation model for histology analysis
• Combines reasoning over clinical and genomic data sources
• Fuses multimodal features to enhance mutation prediction accuracy