MERGE: Multi-faceted Hierarchical Graph-based GNN for Gene Expression Prediction from Whole Slide Histopathology Images

Aniruddha Ganguly,Debolina Chatterjee,Wentao Huang,Jie Zhang,Alisa Yurovsky,Travis Steele Johnson,Chao Chen

2024-12-04

Abstract:Recent advances in Spatial Transcriptomics (ST) pair histology images with spatially resolved gene expression profiles, enabling predictions of gene expression across different tissue locations based on image patches. This opens up new possibilities for enhancing whole slide image (WSI) prediction tasks with localized gene expression. However, existing methods fail to fully leverage the interactions between different tissue locations, which are crucial for accurate joint prediction. To address this, we introduce MERGE (Multi-faceted hiErarchical gRaph for Gene Expressions), which combines a multi-faceted hierarchical graph construction strategy with graph neural networks (GNN) to improve gene expression predictions from WSIs. By clustering tissue image patches based on both spatial and morphological features, and incorporating intra- and inter-cluster edges, our approach fosters interactions between distant tissue locations during GNN learning. As an additional contribution, we evaluate different data smoothing techniques that are necessary to mitigate artifacts in ST data, often caused by technical imperfections. We advocate for adopting gene-aware smoothing methods that are more biologically justified. Experimental results on gene expression prediction show that our GNN method outperforms state-of-the-art techniques across multiple metrics.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is: how to predict gene expression more accurately from Whole Slide Histopathology Images (WSI). Specifically, existing methods fail to fully utilize the interactions between different tissue locations when predicting gene expression, which limits the prediction accuracy. In addition, the noise and sparsity in Spatial Transcriptomics (ST) data also pose challenges to the prediction task.

### Core problems of the paper and solutions

1. **Lack of effective modeling of long - range interactions**:
- Existing methods mainly focus on local or short - range interactions and ignore the interactions between distant but biologically relevant tissue regions.
- Solution: Introduce the MERGE model. By constructing a multi - faceted hierarchical graph and combining it with Graph Neural Networks (GNN), it promotes long - range interactions between different tissue locations. Specific methods include:
- **Spatial clustering**: Cluster according to tissue location coordinates to capture the similarity of adjacent tissue regions.
- **Feature - space clustering**: Cluster according to image features to capture the similarity of morphologically similar but spatially non - adjacent tissue regions.
- **Shortcut edges**: Establish connections between different clusters so that information can spread quickly between distant nodes.

2. **Quality problems of ST data**:
- ST data has noise, sparsity, and a high dropout rate, resulting in inaccurate prediction results.
- Solution: Propose a gene - informed bimodal smoothing technique based on gene information, which combines spatial correlation and gene expression patterns for smoothing processing to improve data quality and prediction accuracy.

### Main contributions of the MERGE model

- **Improved Graph Neural Network structure**: By constructing a multi - faceted hierarchical graph, the MERGE model can better capture local and global dependencies, thereby improving the accuracy of gene expression prediction.
- **Multi - faceted hierarchical graph construction strategy**: Utilize spatial and feature - space clustering to identify and utilize the dependencies between similar tissue clusters, regardless of whether they are spatially close or not.
- **Gene - information - guided smoothing technique**: Adopt a gene - aware smoothing method, which significantly improves the correlation between gene expression values and tissue morphology and enhances the similarity and difference between distant points.

Through these innovations, the MERGE model outperforms existing methods on multiple evaluation metrics, especially in the gene expression prediction task.