Table 2:

PAM and K-means clustering accuracy. Shown for all tree and dual graph topologies with 3 or more vertices are the motifs classified as RNA-like and non RNA-like by unsupervised clustering algorithms PAM and K-means using reduced linear and quadratic variables. Also shown are the number and percentage of existing tree and dual graph topologies correctly classified as RNA-like.

	(a) Tree graphs
	All Topologies (Total: 2286)		Existing Topologies (Total: 79)
	RNA-like	non RNA-like	RNA-like	non RNA-like
Method	Linear variables
PAM	1645	641	61	18
	(71.96%)	(28.04%)	(77.22%)	(22.78%)
K-means	1643	643	61	18
	(71.87%)	(28.13%)	(77.22%)	(22.78%)
Method	Quadratic variables
PAM	1889	397	58	21
	(82.63%)	(17.37%)	(73.42%)	(26.58%)
K-means	1890	396	58	21
	(82.68%)	(17.32%)	(73.42%)	(26.58%)
	(b) Dual graphs
	All Topologies (Total: 110,664)		Existing Topologies (Total: 118)
	RNA-like	non RNA-like	RNA-like	non RNA-like
Method	Linear variables
PAM	55,268	55,396	89	29
	(49.94%)	(50.06%)	(75.42%)	(24.58%)
K-means	55,258	55,406	89	29
	(49.93%)	(50.07%)	(75.42%)	(24.58%)
Method	Quadratic variables
PAM	57,078	53,586	86	32
	(51.58%)	(48.42%)	(72.88%)	(27.12%)
K-means	56,994	53,670	86	32
	(51.50%)	(48.50%)	(72.88%)	(27.12%)