Path2ST: Hierarchical Cell-Tissue Grounded Cross-Modal Translation for Spatial Transcriptomics

2026-08-11

AI Summary

This study addresses the limitations of existing methods in capturing biological hierarchical structures and modeling microenvironmental regulation by proposing Path2ST, a framework that reframes H&E-to-spatial transcriptomics prediction as a cross-modal semantic translation task. The approach introduces a novel hierarchical cell-tissue conditional fusion mechanism, scale-adaptive autoregressive generation, and a SpectraLoss function to enable coarse-to-fine synthesis of biologically consistent gene expression and transcriptional bursting. Experimental evaluations across three datasets demonstrate that Path2ST achieves state-of-the-art performance, significantly enhancing both the accuracy and spatial coherence of generated transcriptomic profiles compared to existing baselines.

Abstract

Predicting spatial gene expression from hematoxylin and eosin (H\&E)-stained images offers a cost-effective alternative to spatial transcriptomics (ST). However, existing methods treat H\&E images as generic visual inputs and ignore their intrinsic biological hierarchy, where spatially organized cell types collectively form functional tissue microenvironments that govern local gene expression programs. To bridge this gap, we formulate H\&E-to-ST prediction as a cross-modal semantic translation task and propose Path2ST, a hierarchically grounded autoregressive framework featuring three key components: (i) a Hierarchical Cell-Tissue Conditioning mechanism that fuses explicit and implicit cellular features with tissue-level semantic representations to construct hierarchical conditioning signals; (ii) a Scale-Adaptive Autoregressive Generation process over a hierarchical semantic vocabulary, enabling coarse-to-fine, biologically consistent expression synthesis; and (iii) SpectraLoss, a full-spectrum objective that jointly enforces ordinal fidelity, models transcriptional bursts, and aligns semantic structures with cell types. Extensive experiments on three datasets demonstrate state-of-the-art performance, validating that Path2ST generates highly accurate and spatially coherent transcriptomic profiles. The related code is released at https://github.com/RuochenLiu23/Path2ST.

Problem

• Spatial Transcriptomics
• H&E-stained Images
• Cross-Modal Translation
• Biological Hierarchy
• Gene Expression Prediction

Innovation

• Hierarchical Cell-Tissue Conditioning
• Scale-Adaptive Autoregressive Generation
• SpectraLoss
• Cross-Modal Semantic Translation
• Spatial Transcriptomics