Efficient and automated large-scale detection of structural relationships in proteins with a flexible aligner

BMC Bioinformatics

Table 1 Performance benchmark analysis of MOMA with different methods

Methods	AUC		ACC		^*fp		^*tp		time
	Fold	Superfamily	Fold	Superfamily	Fold	Superfamily	Fold	Superfamily
Structal	0.956	0.969	0.902	0.919	0.076	0.060	0.880	0.898	10d 21h (1,842x)
TopMatch	0.955	0.974	0.883	0.911	0.121	0.069	0.887	0.891	2d (339x)
MOMA	0.940	0.956	0.872	0.889	0.139	0.113	0.884	0.891	8m 28s (1x)
GANGSTA+	0.916	0.911	0.845	0.851	0.101	0.058	0.791	0.761	5d 6h 49m (895x)
QP tableau search	0.877	0.918	0.791	0.831	0.224	0.188	0.805	0.850	2d 7h 27m (391x)
SHEBA	0.870	0.889	0.841	0.875	0.052	0.042	0.734	0.793	6h 51m (48x)
FATCAT flexible	0.837	0.911	0.743	0.825	0.220	0.211	0.706	0.862	27d 2h 38m (4,614x)
YAKUSA	0.790	0.858	0.727	0.794	0.155	0.088	0.609	0.677	48m (5.7x)

Area under ROC curve (AUC), maximal accuracy (ACC), false positive (fp) and true positive (tp) rates for each method are reported (^*these values are calculated at the same threshold that gives the maximum accuracy reported as ACC). The execution time needed to compare the 100 queries against the 11,121 domains in the ASTRAL SCOP 40% sequence identity dataset is shown in the last column of the table. Execution times are reported in seconds (s), minutes (m), hours (h) and days (d) (in parenthesis, the speed gain factor of MOMA when compared to other methods is displayed, where “x” means number of times faster)

ISSN: 1471-2105