Kohonen Self Organized Map
This is the same figure as depicted in Fig. 4 of the paper, except that it has been expanded along the y-axis and gene names have been added for each row. The method is an "unsupervised" neural-net algorithm which, in this case, attempts to map expression profiles along a one dimensional axis such that closely related data are positioned near to each other. The metric used for being "closely related" is the correlation coefficient.