Skip to main content
. Author manuscript; available in PMC: 2012 Nov 1.
Published in final edited form as: Nat Methods. 2012 Mar 18;9(5):473–476. doi: 10.1038/nmeth.1937

Fig. 1.

Fig. 1

Heat map of discovered Gaussian parameters in an unsupervised 25-label segmentation of 31 ENCODE signal tracks. Each row contains parameters for one signal track, and each column contains parameters for one segment label. Within each row, we performed an affine transformation such that the largest mean was 1 and the smallest 0. The color in each cell indicates the transformed mean parameter μ according to the color bar on the left. The width of the inner boxes is proportional to the square root of the variance parameter σ2, after multiplying by the linear factor used in the transformation of μ. Dendrograms show a hierarchical clustering by both rows and columns. We compared the model parameters and segmentation against existing literature to manually assigned functional categories to the segment labels (D: dead, F: FAIRE, R: repression, H3K9me1: histone 3 lysine 9 monomethylation, L: low, GE: gene end, TF: transcription factors, C: CTCF, TSS: transcription start site, GS: gene start, E: enhancer, GM: gene middle).