Hyperbolic learning improves gene maps from tissue images
Hyperbolic Contrastive Learning with Entailment for Spatial Transcriptomics
Computer Vision and Pattern Recognition
Summary
Spatial transcriptomics lets scientists see where genes are active inside tissues but is costly and sensitive to errors. The authors created a new computer method called HyCLoST that uses hyperbolic geometry to better capture complex relationships between genes and tissue shapes. This method produces more accurate maps of gene activity from tissue pictures by respecting the natural hierarchy in biology. Their approach reduced errors by 6% and improved correlations by 8%, helping predict gene expression patterns more reliably.
What this means in practice
- •For biomedical imaging teams: Predict spatial gene expression from tissue images with higher accuracy to support biological analysis.
- •For pharmaceutical research teams: Improve tissue-based gene activity predictions to aid drug development and disease study without expensive spatial transcriptomics setups.
Authors
Daniela Vega, Paula Cárdenas, Hannah Ceballos, Leonardo Manrique, Pablo Arbelaéz
Abstract
Spatial Transcriptomics (ST) has transformed biomedical research by enabling the spatial mapping of gene expression across tissue sections. However, high operational costs, specialized equipment requirements, and sensitivity to experimental noise limit the accessibility and scalability of ST. Recent computer vision approaches aim to overcome these limitations by predicting spatial gene expression directly from histopathology images. While effective, current approaches often suffer from gene expression over-smoothing and overly uniform predictions across tissue regions, suggesting that further progress depends on learning representations that reflect the hierarchical and asymmetric structure of gene regulation and tissue morphology. To address these issues, we propose Hyperbolic Contrastive Learning with Entailment for Spatial Transcriptomics (HyCLoST), a hyperbolic contrastive learning model that captures the intrinsic hierarchical relationships within ST data. By leveraging hyperbolic geometry and a gene-to-image entailment loss, HyCLoST learns structured, biologically grounded representations that improve gene expression prediction accuracy, achieving a 6% reduction in MSE and an 8% increase in PCC across 26 ST datasets, over previous methods. Our source code is publicly available at https://github.com/BCV-Uniandes/HyCLoST