Tableau-based protein substructure search using quadratic programming

BMC Bioinformatics

Table 1 AUC and time for some widespread folds.

			distance information
			without		with
Fold	SCOP sid	# SSEs	AUC	time	AUC	time
β-grasp	d1ubia_	8	0.80	0 h 47 m	0.92	0 h 32 m
Key-barrel	d1tttb1	9	0.80	0 h 50 m	0.97	0 h 50 m
Immunoglobulin	d1ae6h1	13	0.89	1 h 47 m	0.95	1 h 53 m
Plait (ferredoxin)	d1bhne_	15	0.61	1 h 53 m	0.85	2 h 18 m
GFP-like	d1h6rb_	17	1.00	3 h 06 m	0.99	3 h 06 m
Jelly-roll	d2phlb1	19	0.87	4 h 24 m	0.93	5 h 13 m
Tim-barrel	d1tima_	21	0.99	4 h 31 m	1.00	4 h 30 m
NAD-binding fold	d1f6dc_	30	0.98	14 h 45 m	0.99	16 h 33 m

Area under ROC curve (AUC) and time taken for searches for several structural folding patterns against the ASTRAL SCOP 95% sequence identity non-redundant database consisting of 15273 domains using discrete tableaux, both with and without incorporating SSE distance information. Times (in hours and minutes) reported are elapsed times using the sparse matrix (UMFPACK) implementation on an Intel Xeon 3.2 GHz machine with 8 GB memory running Linux. The table is sorted by query size (number of SSEs in query).

ISSN: 1471-2105