A Multimodal Foundation Model of Spatial Transcriptomics and Histology for Biological Discovery and Clinical Prediction

STORM: Multimodal Foundation Model of Spatial Transcriptomics and Histology.

Abstract

We developed STORM, a foundation model integrating spatial transcriptomics and histology imaging. Trained on approximately 1.2 million transcriptomic profiles across 18 organs, the model bridges molecular and morphological data. Testing across 23 independent patient cohorts showed improved predictions for immunotherapy response and clinical outcomes compared to existing approaches. The system demonstrates compatibility with multiple sequencing platforms including Visium, Xenium, Visium HD, and CosMx.

Publication
arXiv, 2026
Jinxi Xiang
Jinxi Xiang
Postdoctoral Fellow in Medical AI

My research interests include computer vision and medical image analysis.