Cutoff Scanning Matrix (CSM): structural classification and function prediction by protein inter-residue distance patterns

Table 1 Function prediction performance using KNN for the gold-standard dataset

Superfamily	Before SVD		After SVD		∆Prec.	∆Rec.
	Precision	Recall	Precision	Recall
Amidohydrolase	0.983	0.983	1.000	1.000	+1.7%	+1.7%
Crotonase	0.955	0.953	0.979	0.977	+2.4%	+2.4%
Enolase	0.876	0.853	0.971	0.967	+9.5%	+11.4%
Haloacid Dehalogenase	0.881	0.925	0.984	0.981	+10.3%	+5.6%
Isoprenoid Synthase Type I	1.000	1.000	1.000	1.000	+0.0%	+0.0%
Vicinal Oxygen Chelate	1.000	1.000	1.000	1.000	+0.0%	+0.0%
All	0.901	0.903	0.991	0.989	+9.0%	+8.6%

Prediction performance for the gold-standard dataset using KNN. The experiment was performed in an intra-superfamily fashion, and the classes for prediction represent the enzyme’s families. The precision and recall metrics are weighted averages. Ten-fold cross validation was employed.

ISSN: 1471-2164