Researchers from Dartmouth‘s Norris Cotton Cancer Center reported in Pacific Symposium on Biocomputing on the use of denoting autoencoders to effectively extract key biological principles from gene expression data and summarize them into constructed features with convenient properties.