Fig. 8: TP53 genotype heatmaps based on predicted probabilities using our deep learning model. | Nature Communications

Fig. 8: TP53 genotype heatmaps based on predicted probabilities using our deep learning model.

From: Deep learning-based cross-classifications reveal conserved spatial behaviors within tumor histological images

Fig. 8

The first row shows two LUAD H&E slides with TP53 mutant (left panel) and wild type (right panel). The second row shows prediction maps for these two slides using tumor/normal classifiers trained on BRCA and LUAD samples. Both models successfully classify samples as cancer and predict similar heatmaps. The third row shows prediction maps for these slides using TP53 mutation classifiers trained on BRCA and LUAD. The BRCA-trained and LUAD-trained heatmaps are similar, suggesting that there are spatial features for TP53 mutation that are robust across tumor types.

Back to article page