EXAONE Path 2.5: Pathology Foundation Model with Multi-Omics Alignment

2025-12-15

AI Summary

Cancer progression involves multi-scale biological mechanisms that cannot be fully captured by histopathological images alone. To address this, we propose the first multimodal foundation model integrating whole-slide imaging (WSI), genomics, epigenomics, and transcriptomics (RNA-seq) to construct patient-level unified biological representations. Methodologically, we introduce a multimodal SigLIP contrastive loss, fragment-aware rotary position encoding (F-RoPE), and domain-specific foundational modules for WSI and RNA-seq—enabling biology-informed cross-modal alignment. Evaluated on Patho-Bench—a comprehensive benchmark comprising 80 downstream tasks—our model achieves state-of-the-art performance. It demonstrates superior generalizability and adaptability on real-world clinical data, while maintaining high data and parameter efficiency. This work establishes a scalable, biologically grounded framework for integrative computational pathology.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

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

Abstract

Cancer progression arises from interactions across multiple biological layers, especially beyond morphological and across molecular layers that remain invisible to image-only models. To capture this broader biological landscape, we present EXAONE Path 2.5, a pathology foundation model that jointly models histologic, genomic, epigenetic and transcriptomic modalities, producing an integrated patient representation that reflects tumor biology more comprehensively. Our approach incorporates three key components: (1) multimodal SigLIP loss enabling all-pairwise contrastive learning across heterogeneous modalities, (2) a fragment-aware rotary positional encoding (F-RoPE) module that preserves spatial structure and tissue-fragment topology in WSI, and (3) domain-specialized internal foundation models for both WSI and RNA-seq to provide biologically grounded embeddings for robust multimodal alignment. We evaluate EXAONE Path 2.5 against six leading pathology foundation models across two complementary benchmarks: an internal real-world clinical dataset and the Patho-Bench benchmark covering 80 tasks. Our framework demonstrates high data and parameter efficiency, achieving on-par performance with state-of-the-art foundation models on Patho-Bench while exhibiting the highest adaptability in the internal clinical setting. These results highlight the value of biologically informed multimodal design and underscore the potential of integrated genotype-to-phenotype modeling for next-generation precision oncology.

Problem

• Integrates histologic, genomic, epigenetic, transcriptomic data for comprehensive tumor representation
• Enables multimodal contrastive learning and spatial structure preservation in pathology
• Evaluates performance across clinical and benchmark tasks for precision oncology

Innovation

• Multimodal contrastive learning across histologic, genomic, epigenetic, transcriptomic data
• Fragment-aware rotary encoding preserving spatial structure in whole slide images
• Domain-specialized foundation models for robust multimodal biological alignment