Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Spatially Resolved Gene Expression Prediction from Histology via Multi-view Graph Contrastive Learning with HSIC-bottleneck Regularization

Changxi Chi, Hang Shi, Qi Zhu, Daoqiang Zhang, Wei Shao

TL;DR

This work tackles the challenge of predicting spatial gene expression from histology by incorporating spatial context and cross-modal alignment. The authors introduce ST-GCHB, a multi-view graph contrastive learning framework with HSIC-bottleneck regularization to learn shared representations between histology images and spatial transcriptomics data, leveraging both intra- and inter-modal signals. Through a retrieval-based prediction pipeline, ST-GCHB achieves higher correlations with ground-truth expression on the DLPFC/10X Visium dataset, notably surpassing several baselines and highlighting the value of preserving spatial structure while suppressing redundant information via $nHSIC$. The approach demonstrates improved imputation accuracy and the ability to reveal spatial gene expression patterns, supporting scalable, image-driven analysis in spatial genomics.

Abstract

The rapid development of spatial transcriptomics(ST) enables the measurement of gene expression at spatial resolution, making it possible to simultaneously profile the gene expression, spatial locations of spots, and the matched histopathological images. However, the cost for collecting ST data is much higher than acquiring histopathological images, and thus several studies attempt to predict the gene expression on ST by leveraging their corresponding histopathological images. Most of the existing image-based gene prediction models treat the prediction task on each spot of ST data independently, which ignores the spatial dependency among spots. In addition, while the histology images share phenotypic characteristics with the ST data, it is still challenge to extract such common information to help align paired image and expression representations. To address the above issues, we propose a Multi-view Graph Contrastive Learning framework with HSIC-bottleneck Regularization(ST-GCHB) aiming at learning shared representation to help impute the gene expression of the queried imagingspots by considering their spatial dependency.

Spatially Resolved Gene Expression Prediction from Histology via Multi-view Graph Contrastive Learning with HSIC-bottleneck Regularization

TL;DR

This work tackles the challenge of predicting spatial gene expression from histology by incorporating spatial context and cross-modal alignment. The authors introduce ST-GCHB, a multi-view graph contrastive learning framework with HSIC-bottleneck regularization to learn shared representations between histology images and spatial transcriptomics data, leveraging both intra- and inter-modal signals. Through a retrieval-based prediction pipeline, ST-GCHB achieves higher correlations with ground-truth expression on the DLPFC/10X Visium dataset, notably surpassing several baselines and highlighting the value of preserving spatial structure while suppressing redundant information via . The approach demonstrates improved imputation accuracy and the ability to reveal spatial gene expression patterns, supporting scalable, image-driven analysis in spatial genomics.

Abstract

The rapid development of spatial transcriptomics(ST) enables the measurement of gene expression at spatial resolution, making it possible to simultaneously profile the gene expression, spatial locations of spots, and the matched histopathological images. However, the cost for collecting ST data is much higher than acquiring histopathological images, and thus several studies attempt to predict the gene expression on ST by leveraging their corresponding histopathological images. Most of the existing image-based gene prediction models treat the prediction task on each spot of ST data independently, which ignores the spatial dependency among spots. In addition, while the histology images share phenotypic characteristics with the ST data, it is still challenge to extract such common information to help align paired image and expression representations. To address the above issues, we propose a Multi-view Graph Contrastive Learning framework with HSIC-bottleneck Regularization(ST-GCHB) aiming at learning shared representation to help impute the gene expression of the queried imagingspots by considering their spatial dependency.
Paper Structure (13 sections, 7 equations, 2 figures, 2 tables, 1 algorithm)

This paper contains 13 sections, 7 equations, 2 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Method
  3. Dataset
  4. Raw Data Preprocessing
  5. Spatial Adjacency Information
  6. HSIC-Bottleneck
  7. Contrastive Learning
  8. Prediction
  9. Experiments and Results
  10. Correlation of Expression Prediction
  11. Ablation Study
  12. Experiment Settings
  13. Conclusion

Figures (2)

  • Figure 1: ST-GCHB extracts features from the STs gene expression and adjacency structure of sampling points, incorporating the nHSIC-Bottleneck structure to further eliminate redundant information. Simultaneously, it segments corresponding image patches from H&E stained images to extract image features. Ultimately, ST-GCHB aligns matching feature pairs in the feature space.
  • Figure 2: Visualization in spatial coordinates obtained from four methods.