A method for probabilistic mapping between protein structure and function taxonomies through cross training

BMC Structural Biology

Table 2 Performance evaluation for some functional classes

Class Name	FP Rate	Precision	Recall	F-measure
Immunoglobulins and MHC protein	0.002	0.765	0.619	0.684
Ig-like domain profile	0.005	0.547	0.343	0.422
Cytochome c family, heme-binding site signature	0	0.988	0.908	0.946
Globins family	0.002	0.792	0.623	0.697
Pyridine nucleotide-disulphide oxidoreductases c-I	0.001	0.704	0.559	0.623
Serine protease trypsin family	0.001	0.918	0.789	0.848
ATP dependent helicases signatures	0.002	0.52	0.419	0.464
Phospholipase A2 active site signature	0	1	0.862	0.926
Nuclear hormones receptors DNA-binding	0.003	0.667	0.561	0.61

Performance evaluation while trying to classify functional classes using structural classes as the features. The high F-measure values (Equation 1) indicate that the function is strongly dependent on the structure that the protein has. F-measure was calculated as the harmonic mean between Recall and Precision. Class-name refers to domain in PROSITE database. Highly significant F-measure values are shown in bold.

ISSN: 1472-6807