GC-MoE: Genomics-Guided Cell-Type-Specific Mixture of Experts for Histology-Based Single-Cell Spatial Transcriptomics

2026-06-01

AI Summary

This work proposes a novel computational method to predict single-cell gene expression from histopathology images and cell coordinates as a cost-effective alternative to expensive spatial single-cell transcriptomics. The approach introduces, for the first time, a cell-type-guided mixture-of-experts (MoE) architecture: a deep routing network estimates cell-type probabilities to softly combine outputs from cell-type-specific expert models. It further integrates a co-expression-aware predictor (CAP) with a lightweight cell-to-cell attention (C2CA) module to effectively model both cell-type-dependent gene programs and local microenvironmental influences. Evaluated on multiple public datasets, the method significantly outperforms existing single-cell and spot-level baselines, demonstrating superior prediction accuracy and generalization capability.

Abstract

Histology-based single-cell spatial transcriptomics (ST) estimation aims to predict gene expression for individual cells from histopathological images and cell locations, reducing the need for costly single-cell ST measurements. Unlike existing histology-to-ST methods that mainly predict spot-level profiles for local regions containing multiple cells, this task requires modeling cell-to-cell expression variability, which is strongly structured by cell type. We propose Genomics-Guided Cell-Type-Specific Mixture-of-Experts (GC-MoE), which estimates cell-type probabilities with a routing network and softly combines cell-type-specific experts for gene expression prediction. To further encode cell-type-dependent gene programs, we introduce the Cell-Type-Specific Co-Expression-Aware Predictor (CAP), together with a lightweight Cell-to-Cell Interaction Attention (C2CA) module for neighboring-cell context. Experiments and ablations on public single-cell ST datasets show consistent improvements over existing single-cell and adapted spot-level baselines.

Problem

• spatial transcriptomics
• single-cell
• histology
• cell-type-specific
• gene expression prediction

Innovation

• Mixture-of-Experts
• Cell-Type-Specific Modeling
• Spatial Transcriptomics
• Co-Expression-Aware Prediction
• Cell-to-Cell Interaction Attention