Fig. 3
From: N-of-one differential gene expression without control samples using a deep generative model
Closest normal representation for cancer samples. A Schematic representation of the procedure to find closest-normal representations for TCGA tumors. We find the representation by maximizing the probability of the representation while leaving the GMM and neural network fixed. B Density plot showing the negative log probability distributions for GTEx test, TCGA-normal, and TCGA tumor samples. C Matrix of associations showing the percentage of TCGA tumor samples assigned to GMM components representative of tissues. GMM components are named by their majority normal tissue type (see Fig. 2B)