Fig. 4: Validation of the super-resolved expression maps.
From: A deep learning-based multiscale integration of spatial omics with tumor morphology
a H&E sample of colorectal cancer. b. Ground truth expression of IGKC obtained with the 10X Xenium technology. For comparison with the model’s predictions, transcripts were summed on patches of size 7 µm x 7 µm. Brighter shades indicate higher expression c. Prediction of the expression of IGKC by the MISO super-resolution approach. d, e Same as b. and c., for the expression of EPCAM. f, g Ground truth and predicted expression of EPCAM, zoomed in in the area indicated by the blue squares in d. and e. h. H&E sample of breast carcinoma. i Ground truth expression of EPCAM obtained with the 10X Xenium technology. j Prediction of the expression of EPCAM by the MISO super-resolution approach. Scale bars in (a−g). 2 mm.
