Extended Data Fig. 10: Results from analyses of a DLVPM model trained on TCGA data and applied to spatial transcriptomics data.
From: Integrating multimodal cancer data using deep latent variable path modelling
Tile-wise heatmaps generated from the DLVPM model, trained on TCGA data, and applied to histological and associated spatial transcriptomic data. The colormap is flipped for the histology heatmap as this DLV shows a negative association with the genes of interest. The association/significance matrices on the right show correlations between genes of interest and the first histology DLV for both tumours. The upper triangular part of each matrix is denoted with Pearson’s Correlation Coefficient between each gene, and the histology data. The lower triangular part of each matrix denotes the significance level between genes and histology data.
