Designing and benchmarking the MULTICOM protein structure prediction system

BMC Structural Biology

Table 1 The specificity and sensitivity for the top 1 template and the best template among the top 5 templates for each tool based on 107 CASP9 targets and the number of targets that have templates identified for each tool

Tool	The top 1 model		The best model		# targets that have templates identified
	Specificity	Sensitivity	Specificity	Sensitivity
PSI-BLAST-multi	80.28%	53.27%	88.73%	58.88%	71
CS-BLAST	73.97%	50.47%	78.08%	53.27%	73
CENTER	67.29%	67.29%	71.96%	71.96%	107
STAR	67.29%	67.29%	71.96%	71.96%	107
HMMER	66.67%	56.07%	77.78%	65.42%	90
SS	66.04%	65.42%	71.96%	71.96%	107
HHSearch	65.42%	65.42%	72.90%	72.90%	107
BLAST	65.38%	47.66%	69.23%	50.47%	78
CSI-BLAST	62.63%	57.94%	66.67%	61.68%	99
COMPASS	62.50%	60.75%	71.15%	69.16%	104
PSI-BLAST-single	62.50%	56.07%	67.71%	60.75%	96
PRC	62.14%	59.81%	69.90%	67.29%	103
SAM	61.32%	60.75%	67.92%	67.29%	106
CONSTRUCT	60.75%	60.75%	71.96%	71.96%	107

The specificity is the fraction of the targets with at least one template identified by a method having a GDT-TS score > = 0.5, i.e. the number of targets for which a good template is identified divided by the number of targets for which at least one template is identified. The specificity measures the precision of template identification of a method. The sensitivity is the number of targets for which a good template (i.e. its GDT-TS score >0.5) is identified divided by all the targets in consideration in this experiment (i.e. 107), assuming that all the targets have at least one reasonable template. The two measures (i.e. sensitivity and specificity) are complementary.

ISSN: 1472-6807