Prior Knowledge Injection into Deep Learning Models Predicting Gene Expression from Whole Slide Images

2025-01-23

AI Summary

Predicting genome-wide gene expression profiles solely from whole-slide histopathology images (WSIs) suffers from low accuracy and poor generalizability. Method: We propose a model-agnostic prior knowledge injection framework that integrates gene–gene interaction networks into WSI analysis in a plug-and-play manner. Our approach combines graph neural network embeddings, attention-guided knowledge fusion, and multi-scale feature extraction (using ResNet and Vision Transformers), trained and validated on TCGA-BRCA data. Contribution/Results: This work is the first to systematically demonstrate that incorporating gene regulatory priors significantly enhances generalization in expression prediction. In breast cancer, it yields an average of 983 additional significantly predicted genes (p < 0.05) per experiment; 14 out of 18 experiments show consistent performance on independent test sets, and cross-architecture prediction robustness is markedly improved.

Technology Category

• Computer Vision: Other Foundations of Computer Vision
• Reasoning under Uncertainty: Graphical Models
• Machine Learning: Graph-based Machine Learning

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Cancer diagnosis and prognosis primarily depend on clinical parameters such as age and tumor grade, and are increasingly complemented by molecular data, such as gene expression, from tumor sequencing. However, sequencing is costly and delays oncology workflows. Recent advances in Deep Learning allow to predict molecular information from morphological features within Whole Slide Images (WSIs), offering a cost-effective proxy of the molecular markers. While promising, current methods lack the robustness to fully replace direct sequencing. Here we aim to improve existing methods by introducing a model-agnostic framework that allows to inject prior knowledge on gene-gene interactions into Deep Learning architectures, thereby increasing accuracy and robustness. We design the framework to be generic and flexibly adaptable to a wide range of architectures. In a case study on breast cancer, our strategy leads to an average increase of 983 significant genes (out of 25,761) across all 18 experiments, with 14 generalizing to an increase on an independent dataset. Our findings reveal a high potential for injection of prior knowledge to increase gene expression prediction performance from WSIs across a wide range of architectures.

Problem

• Deep Learning
• Gene Activity Prediction
• Image-based Molecular Information

Innovation

• Gene Expression Prediction
• Deep Learning with Prior Knowledge
• Breast Cancer Research