Figure 7.

PCA of metacluster terms from GREAT identifies groups with different functional associations. The two first components from a Principal Component Analysis (PCA) on the functional terms from GREAT for all clusters in all cell lines. Cell line abbreviations are K5 for K562, Gm for GM12878, He for HeLa-S3 and Hp for HepG2, and c1-c10 are cluster indexes as used previously. A) PCA plot for all cluster scores. Clusters with scores in the same quadrant of the PCA plot have related functional terms, and the distance between two clusters corresponds to the degree of similarity. Clusters with scores close to origo (the crossing of the axes arrows) have few functional terms associated with them, while those with scores further away from origo are enriched for several terms. B) Interpretations of groupings in the PCA plot, with additional cluster properties also taken into consideration. Three main groups are apparent, with one group (yellow shading) also displaying within-group variation. The terms proximal, normal and distal refers to the localization of RTSS enrichment relative to genes as calculated by GREAT (Figure 6). Interestingly, the more distal clusters with low or intermediate RTSS expression show the strongest functional associations. Displayed PCA plot are from the permissive GREAT analysis. A PCA plot from the conservative analysis are in [Additional file 1: Figure S9].