📝 Preprints

BRIDGE: A Cross-organ Foundation Model for Bridging Histology Imaging and Spatial Transcriptomics
Zhuo Liang$^{*}$, Weiqin Zhao$^{*}$, Fuying Wang, Yuanhua Huang, Lequan Yu.

Nature Communications Under Review   

Present a multi-organ foundation model pre-trained on over 600,000 image-gene paired profiles from 13 organs with various sequencing techniques, to connect cross-organ morphology and genomics for cost-effective histology analysis. Without fine-tuning, BRIDGE enables accurate gene inference for over 80 biomarker genes across 10 human organs, enhances survival analysis for six different types of cancer by using histopathology images from TCGA cohorts and shows potential as a cost-efficient alternative to bulk RNA sequencing.

Knowledge-Guided Adaptation of Pathology Foundation Models Effectively Improves Cross-domain Generalization and Demographic Fairness
Yanyan Huang, Weiqin Zhao, Yihang Chen, Yu Fu, Yuming Jiang, Li Liang, Shujun Wang, Lequan Yu.

Nature Communications Under Revision   

Proposed FLEX, a novel framework that enhances pathology foundation models to address critical limitations in cross-domain generalization and demographic fairness caused by site-specific signatures and biases. FLEX utilizes a task-specific information bottleneck guided by visual and textual domain knowledge to align features across diverse clinical sites and demographic groups. Evaluated on 16 tasks across 4 TCGA cohorts, FLEX significantly improves out-of-distribution performance, enhances fairness, and shows high versatility with different VLMs and MIL methods, promoting more robust and equitable computational pathology.