Spatial Transcriptomics-Guided Alignment Enhances Molecular Profiling in Pathology Foundation Model

2026-05-29

AI Summary

Current foundation models in computational pathology struggle to accurately link histological morphology with genomic alterations due to the absence of spatially resolved molecular supervision, thereby limiting their ability to infer molecular phenotypes directly from H&E-stained slides. To address this, this work proposes STAMP, a novel framework that introduces, for the first time, a spatial transcriptomics–guided molecular alignment mechanism. By integrating pathway-informed representation alignment with parameter-efficient fine-tuning, STAMP endows the model with intrinsic molecular awareness. The study also constructs HumanST-1k, a large-scale multi-organ dataset comprising 1.8 million paired H&E and spatial transcriptomics samples, which substantially enhances the model’s accuracy and clinical applicability in predicting molecular features without requiring additional sequencing.

Abstract

Comprehensive molecular profiling is essential for modern precision oncology but remains hindered by prohibitive costs, specimen exhaustion, and protracted turnaround times. While pathology foundation models (PFMs) have demonstrated potential for inferring molecular phenotypes from routine hematoxylin and eosin (H&E) whole-slide images (WSIs), current architectures primarily rely on vision-centric self-supervised learning or vision-language alignment, lacking the spatially resolved molecular supervision required to connect subtle morphological features with underlying genomic alterations. Spatial transcriptomics (ST) emerges as a transformative technology that enables transcriptomic quantification within intact tissue sections, thereby preserving the precise spatial link between histology and molecular profiles. In this study, we present a Spatial Transcriptomics-guided Alignment framework for Molecular Profiling (STAMP), which endows PFMs with intrinsic molecular awareness. To support this paradigm, we curated HumanST-1k, a human ST dataset spanning diverse anatomical organs and sequencing platforms. This atlas yields 1.8 million pairs of H&E patches and corresponding transcriptomic profiles, providing a corpus that links histological structures with their molecular states. To mitigate the technical noise inherent to raw transcriptomics, STAMP applies a pathway-informed alignment strategy that aggregates transcriptomic data into biologically functional pathways, which are subsequently integrated into PFMs via parameter-efficient fine-tuning. This alignment enriches the representation space of PFMs and unlocks their capacity to resolve sub-visual molecular signatures. The clinical utility of these augmented representations was validated through a multi-tier evaluation framework.

Problem

• Spatial Transcriptomics
• Pathology Foundation Models
• Molecular Profiling
• Histology-Molecular Alignment
• Precision Oncology

Innovation

• Spatial Transcriptomics
• Pathology Foundation Model
• Molecular Profiling
• Pathway-informed Alignment
• HumanST-1k