An adaptive bin framework search method for a beta-sheet protein homopolymer model

BMC Bioinformatics

Table 2 Performance differences among algorithms for the homopolymers of length 12, 24, 32.

Method	Length	E _avg ± sd	E _min	CPU Time _avg	Time _med	Time _{q 75}	Time _{q 25}	p – value
MC	12	-39 (± 0)	-39	< 1 sec	< 1 sec	< 1 sec	< 1 sec
REMC	12	-39 (± 0)	-39	< 1 sec	< 1 sec	< 1 sec	< 1 sec
PHAT	12	-39 (± 0)	-39	< 1 sec	< 1 sec	< 1 sec	< 1 sec
BINMC	12	-39 (± 0)	-39	< 1 sec	< 1 sec	< 1 sec	< 1 sec
MC	24	-109 (± 0)	-109	5.0 min (± 4.1 min)	5.5 min	7.2 min	1.5 min	0.1230
REMC	24	-109 (± 0)	-109	18.3 min (± 18.0 min)	16.3 min	19.4 min	4.3 min	0.0015*
PHAT	24	-109 (± 0)	-109	8.7 min (± 8.2 min)	6.6 min	11.7 min	2.9 min	0.0039*
BINMC	24	-109 (± 0)	-109	1.7 min (± 1.2 min)	1.8 min	2.7 min	0.5 min
Method	Length	E _avg ± sd	Lowest E	CPU Time _avg	E _med	E _{q 75}	E _{q 25}	p – value
MC	32	-158.1 (± 0.9)	-161	4.3 min (± 8.4 min)	-158	-158	-158	0.0155*
REMC	32	-158.2 (± 0.7)	-161	4.5 min (± 8.6 min)	-158	-158	-158	0.0185*
PHAT	32	-158.3 (± 0.9)	-161	5.8 min (± 8.0 min)	-158	-158	-158	0.0214*
BINMC	32	-158.9 (± 0.6)	-161	23.0 min (± 20.1 min)	-159	-159	-158

Comparison of the average energy level obtained and the average time required for the homopolymers of lengths N = 12, 24, 32 for the re-implemented MC, REMC, PHAT, and BINMC. The time cut-off used was 1 CPU hour on our 2.4 GHz reference machine, and all statistics were calculated from 10 independent runs. The temperature sets used for our implementations of MC, REMC and are the same as in Table 1. The p-values reported in the last column were determined using the Mann-Whitney U test to test the null hypothesis that the mean run-time (for N = 24) and the mean energies reached by the respective algorithm vs BINMC (within the same CPU cut-off time for N = 32), respectively, are identical [4]; p-values marked with an asterisk (*) correspond to cases in which the null hypothesis is rejected at a standard significance level of 0.05, and therefore indicate statistically significant performance differences.

ISSN: 1471-2105