DriftST: One-Step Generative Inference of Spatial Transcriptomics from H\&E Histology

2026-07-06

AI Summary

Current methods for inferring spatial gene expression from H&E images suffer from low generation efficiency, neglect of inter-gene dependencies, and difficulty in handling multi-resolution data within a unified framework. To address these limitations, this work proposes DriftST, a novel approach featuring a one-step Cellular Drifting generative model that enables efficient and high-fidelity expression reconstruction. DriftST introduces an STransformer architecture that integrates co-expression attention and gene residual gating to explicitly model gene–gene dependencies and differential importance. Furthermore, it employs a universal gene panel representation capable of supporting both spot-level and cell-level spatial transcriptomics inference in a consistent manner. Extensive experiments demonstrate that DriftST consistently outperforms existing regression-based and multi-step generative methods across diverse tissue types and platforms, achieving state-of-the-art performance.

Abstract

Spatial Transcriptomics (ST) measures gene expression while preserving spatial context, but its high cost and low throughput leave public datasets small. Inferring expression directly from widely available Hematoxylin and Eosin (H&E) stained histology offers a cost-effective alternative. However, existing approaches face several limitations: regression methods over-smooth toward the conditional mean, while generative methods are faithful but require slow multi-step inference; most methods treat genes as independent and equally important, ignoring inter-gene dependencies and heterogeneous gene informativeness; and most are tailored to a single resolution, either spot-level or cell-level. To address these issues, we propose DriftST, a unified framework for inferring spatially resolved gene expression from H&E images. DriftST builds on a Cellular Drifting generative model that learns a direct drift from a histology-conditioned source to the expression distribution, retaining generative expressiveness while enabling efficient one-step generation. To capture gene structure, we introduce the STransformer, which combines a co-expression attention module for inter-gene dependencies with a gene residual gate for differential gene importance. Operating on a generic gene-panel representation, DriftST applies directly to both spot-level and cell-level data in one framework, and extensive experiments across diverse tissues and platforms show that it achieves state-of-the-art performance at both resolutions.

Problem

• Spatial Transcriptomics
• H&E Histology
• Generative Inference
• Gene Expression
• Multi-resolution

Innovation

• DriftST
• Spatial Transcriptomics
• One-step Generative Inference
• STransformer
• Cellular Drifting