Large scale transcriptome data integration across multiple tissues to decipher stem cell signatures.

Summary

A wide variety of stem cells has been reported to exist and renew several adult tissues, raising the question of the existence of a stemness signature-that is, a common molecular program of differentiation. To detect such a signature, we applied a data integration algorithm on several DNA microarray datasets generated by the Stem Cell Genome Anatomy Project (SCGAP) Consortium on several mouse and human tissues, to generate a cross-organism compendium that we submitted to a single layer artificial neural network (ANN) trained to attribute differentiation labels-from totipotent stem cells to differentiated ones (five labels in total were used). The inherent architecture of the system allowed studing the biology behind stem cells differentiation stages and the ANN isolated a 63 gene stemness signature. This chapter presents technological details on DNA microarray integration, ANN training through leave-one-out cross-validation, and independent testing on uncharacterized adult tissues by automated detection of differentiation capabilities on human prostate and mouse stomach progenitors. All scripts of the Stem Cell Analysis and characterization by Neural Networks (SCANN) project are available on the SourceForge Web site: http://scann.sourceforge.net.