Fig. 7
From: Selecting precise reference normal tissue samples for cancer research using a deep learning approach
Comparing signatures from multiple methods. Auto.TopSite: choose all samples from the same tissue of origin based on autoencoder choice, Auto.25 and VarGenes.25: choose 25th percentile and above mostly correlated samples from the same tissue of origin as computed by autoencoder and varying genes, and Auto.Top50 and VarGenes.Top50: choose top 50 mostly correlated samples from all tissues as computed by autoencoder and varying genes. RandomAve: Randomly select 50 samples from all tissues. Site NA means no specified site