Benchmarking the translational potential of spatial gene expression prediction from histology.

Abstract

Spatial transcriptomics has enabled the quantification of gene expression at spatial coordinates across a tissue, offering crucial insights into molecular underpinnings of diseases. In light of this, several methods predicting spatial gene expression from paired histology images have provided the opportunity to enhance the utility of obtainable and cost-effective haematoxylin-and-eosin-stained histology images. To this end, we conduct a comprehensive benchmarking study encompassing eleven methods for predicting spatial gene expression with histology images. These methods are reproduced and evaluated using five Spatially Resolved Transcriptomics datasets, followed by external validation using The Cancer Genome Atlas data. Our evaluation incorporates diverse metrics which capture the performance of predicted gene expression, model generalisability, translational potential, usability and computational efficiency of each method. Our findings demonstrate the capacity of the methods to predict spatial gene expression from histology and highlight areas that can be addressed to support the advancement of this emerging field.