Skip to main content

Table 1 Performance differences among algorithms for the homopolymer of length 64.

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

Method Temperature set Time cut-off E avg ± sd E min P – value
MCSA [9] annealed from 2.75 to 1.25 24 min (approx) -349.3 (± 2.1) -362  
REMC [9] linear 1.25 to 2.75 28 min (approx) -368.2 (± 0.8) -373  
PHAT [10] linear 1.25 to 2.75 1 hr 25 min (approx) -380.4 (± 1.9) -387  
our MC 1.25 24 min -367.2 (± 1.7) -370  
our MC 1.25 28 min -367.4 (± 2.7) -371 0.1367
our REMC linear 1.25 to 2.75 28 min -368.5 (± 2.1) -373 0.3425
our PHAT linear 1.3 to 2.75 28 min -367.5 (± 3.3) -372 0.1599
our BINMC T MC = 1.25, T bin = 6.521 28 min -370.3 (± 4.3) -379  
our MC 1.25 1 hr 25 min -368.2 (± 4.6) -374 0.0006*
our REMC linear 1.25 to 2.75 1 hr 25 min -369.4 (± 3.0) -376 0.0008*
our PHAT linear 1.3 to 2.75 1 hr 25 min -369.5 (± 3.2) -376 0.0023*
our BINMC T MC = 1.25, T bin = 6.521 1 hr 25 min -375.7 (± 3.8) -383  
  1. Comparison of the energy levels reached for the homopolymer of length N = 64 by Monte Carlo Simulated Annealing (MCSA) [9], the Replica Exchange Monte Carlo (REMC) algorithm with a linear set of temperatures [9] and the Parallel-hat Tempering algorithm (PHAT) [10] with our implementation of Monte Carlo (MC), REMC and PHAT, as well as with our new Bin Framework Monte Carlo (BINMC) algorithm. The run-time reported for MCSA and REMC [9] has been conservatively adjusted to our 2.4 GHz reference machine (this was done by dividing the published run-times, which have been obtained on a 500 MHz CPU, by a factor of 4.8). The same was done for the run-times for PHAT (which were originally obtained on a 750 MHz CPU and therefore conservatively divided by 3.2).
\