Evaluating Deep Regression Models for WSI-Based Gene-Expression Prediction

TL;DR

This paper evaluates deep learning regression models for predicting gene expression from whole-slide images, providing insights and recommendations for model training strategies in computational pathology.

Contribution

It offers a comprehensive analysis of training strategies for deep regression models in WSI-based gene-expression prediction, including the effectiveness of a single multi-gene model.

Findings

Training a single model for all genes is computationally efficient.

Deep regression models can effectively predict high-dimensional gene expression.

Recommendations for model training improve prediction performance.

Abstract

Prediction of mRNA gene-expression profiles directly from routine whole-slide images (WSIs) using deep learning models could potentially offer cost-effective and widely accessible molecular phenotyping. While such WSI-based gene-expression prediction models have recently emerged within computational pathology, the high-dimensional nature of the corresponding regression problem offers numerous design choices which remain to be analyzed in detail. This study provides recommendations on how deep regression models should be trained for WSI-based gene-expression prediction. For example, we conclude that training a single model to simultaneously regress all 20530 genes is a computationally efficient yet very strong baseline.