Abstract 1457: Prediction of gene expression and molecular pathway activity from H&E whole slide images in non-small cell lung cancer.

Key Points

• The research aims to predict gene expression and molecular pathway activity from H&E whole slide images in non-small cell lung cancer.
• Developed a two-stage framework for feature extraction using Gigapath.
• Divided H&E images into 256×256 pixel patches and aggregated features using an attention mechanism.
• Evaluated regression models to predict gene expression and pathway activity.
• Used a dataset of 67 NSCLC samples with matched RNA-seq for model evaluation.
• Gigapath-Random Forest regressor showed strong performance with Spearman correlations up to 0.70.
• Identified distinct pathways critical to NSCLC biology with identifiable histological signatures.
• Predicted expression levels of 2,223 genes with a Spearman correlation greater than 0.4.