Gene Ontology-Guided Hierarchical Spatial Gene Expression Prediction from Histopathology Images

2026-07-31

AI Summary

Existing gene expression prediction methods treat genes as unstructured vectors, thereby ignoring the dependencies among genes dictated by biological pathways and regulatory programs, which limits their performance. This work proposes the Multi-Scale Gene Refiner (MSGR), which, for the first time, explicitly incorporates Gene Ontology (GO) as a structural prior into spatial expression prediction. MSGR constructs a four-level functional hierarchy tree and progressively refines predictions during decoding—from coarse-grained functional domains to fine-grained individual genes. Through hierarchical residual correction and scale-weighted supervision, MSGR overcomes the limitations of conventional flat decoding and integrates seamlessly into existing architectures in a plug-and-play manner. Evaluated on nine datasets from the HEST-1k benchmark, MSGR significantly outperforms current methods, achieving a 0.027 improvement over random hierarchical structures, thereby demonstrating the critical contribution of biologically informed ontological structure to predictive accuracy.

Abstract

Predicting spatial gene expression from histopathology images enables large-scale transcriptomic profiling without the cost of direct measurement. Existing methods decode the target gene set as a flat, unstructured vector, ignoring the inter-gene dependencies arising from shared biological pathways and regulatory programs. Without explicit structural guidance, models must infer these dependencies entirely from limited paired data, constraining prediction quality. We propose MSGR (Multi-Scale Gene Refiner), which bridges this gap by incorporating the Gene Ontology (GO), a curated functional hierarchy of genes, as an explicit structural prior. MSGR organizes target genes into a four-level GO tree. Its GO-guided decoder then progressively refines predictions from coarse functional domains to fine individual genes via residual corrections under scale-weighted supervision. Operating solely on the gene side, the GO-guided decoder serves as a seamless plug-in replacement that consistently improves existing architectures without requiring any image-side modifications. Extensive experiments on nine datasets from the HEST-1k benchmark provide empirical evidence for two central claims: GO-structured decoding consistently outperforms flat decoding, even against a state-of-the-art generative baseline, and the gain is attributable to biological ontology structure rather than hierarchical decomposition per se, as confirmed by a +0.027 margin over a structurally equivalent random hierarchy.

Problem

• spatial gene expression
• histopathology images
• Gene Ontology
• gene dependencies
• structured prediction

Innovation

• Gene Ontology
• Hierarchical Prediction
• Spatial Gene Expression
• Histopathology Images
• Multi-Scale Refinement