Table 2:

Random forest classifier performance discriminating high- from lower-confidence predicted protein structures

			Hold-out data Predicted			All data Predicted
			HC	LC	Class error	HC	LC	Class error
All metrics	Actual	HC	228	22	0.088	1054	34	0.031
		LC	24	226	0.096	305	3437	0.082
% AA identity omitted	Actual	HC	227	23	0.092	1048	40	0.037
		LC	29	221	0.116	349	3393	0.093

Metrics for 71 mainly ribosomal protein structures were insufficient for inclusion in data sets for the random forest.