Related Experiment Video
Updated: May 4, 2026
09:19
Spatial Profiling of Protein and RNA Expression in Tissue: An Approach to Fine-Tune Virtual Microdissection
Published on: July 6, 2022
5.5K
Img2Gene: Debiased Spatially Resolved Transcriptomics with Biological Context from Pathology Images.
IEEE Journal of Biomedical and Health Informatics
|April 7, 2026
Summary
Img2Gene accurately predicts gene expression from pathology images by integrating biological context. This debiased framework enhances spatial transcriptomics analysis by addressing data sparsity and improving prediction accuracy.
Area of Science:
- Computational Biology
- Bioinformatics
- Genomics
Background:
- Spatial transcriptomics combines imaging and gene expression data for high-resolution tissue analysis.
- Challenges include data heterogeneity, sparsity, and preserving tissue architecture.
Purpose of the Study:
- To develop Img2Gene, a debiased framework for predicting gene expression from whole slide images.
- To improve the accuracy and reduce bias in spatial gene expression prediction models.
Main Methods:
- Integrating causal analysis to mitigate data sparsity and ensure unbiased predictions.
- Utilizing gene set enrichment analysis for biological context.
- Employing a cross-modal coherence loss to align image and gene expression data.
Main Results:
- Img2Gene achieves state-of-the-art performance across four public datasets.
- Demonstrated improved accuracy in gene expression prediction.
- Enhanced interplay between diverse data features.
Conclusions:
- Img2Gene offers a robust solution for accurate spatial gene expression prediction.
- The framework effectively addresses challenges in integrating imaging and transcriptomic data.
- Provides valuable insights for computational biology and precision medicine.
