Prediction of enzyme function based on 3D templates of evolutionarily important amino acids

BMC Bioinformatics

Table 1 PDB set annotation performance with all matches

	ETA	BLAST	PSI-BLAST	BLAST + ETA	PSI-BLAST + ETA	PEAS (voting)	PEAS (scoring)	PRT (voting)	PRT (scoring)
Matches	164	292	365	122	134	18	18	211	49
True matches	102	186	225	93	101	13	13	53	21
Proteins	49	49	49	49	49	49	49	49	49
With matches	39	44	46	36	36	12	12	49	49
With at least one true match	33	39	40	33	33	9	9	30	21
With vote winners	32	40	39	33	33	10	12	30	49
With correct winners	30	39	37	33	33	7	8	28	21
Prediction accuracy	94%	98%	95%	100%	100%	70%	67%	93%	43%
Prediction availability	65%	82%	80%	67%	67%	20%	24%	61%	100%

PDB set annotation performance from the methods: ETA, BLAST, PSI-BLAST, the BLAST+ETA, PSI-BLAST+ETA, and PEAS using either voting or scoring to pick the most likely function. Prediction accuracy and prediction availability refer to the voting procedure: prediction accuracy = true predictions (i.e., correct vote winners)/total predictions (i.e., vote winners), while prediction availability = total predictions (vote winners)/total number of proteins in that set.

ISSN: 1471-2105