During each constant temperature cycle of Monte Carlo simulated annealing, random changes are made to the ligand's current position, orientation, and conformation, if flexibile. The new state is then compared to its predecesso. If its new energy is lower than the previous, this new state is immediately accepted. However, if the new state's energy is higher than the last, it is accepted probabilistically. This probability depends upon the energy and cycle temperature (see the first equation in Section 2). Generally speaking, at high temperatures, many states will be accepted, while at low temperatures, the majority of these probabilistic moves will be rejected.
The user can choose whether to select the minimum energy state found during a cycle to be used as the initial state for the next cycle, or the last state. The best docking results tend to be achieved by selecting the minimum energy state from the previous cycle.
The initial annealing temperature "
rt0
" should be of the order of the average DE found during the first cycle. This ensures that the ratio of accepted to rejected steps is high at the start. A typical automated docking job may have an initial annealing temperature "
rt0
" of 500 (depending on the system's average DE) and a temperature reduction factor "
rtrf
" of 0.85-0.95 /cycle. Gradual cooling is recommended, to avoid "
simulated quenching
", which tends to trap systems in local minima.
Depending on the degree of complexity of the problem, a relatively good search is given by 50
Monte Carlo
"
cycles
", and a maximum of 30,000 steps rejected "
rejs
" or 30,000 steps accepted "
accs
". 10 "
runs
" may or may not give a range of possible binding modes,. Multiple runs also give relative energies. A schedule of 100 runs, 50 cycles, 3,000 steps accepted, 3,000 steps rejected will provide more highly populated clusters, hinting at the `density of states' for a given conformation. A short test job would be: 1 run, 50 cycles, 100 accepted, 100 rejected steps.
The user must specify the maximum step a state variable can make in one step. Furthermore, these can be adjusted during Monte Carlo simulated annealing, if a reduction factor (a fraction from 0 to less than 1) for translations and rotations is given. At the start of each cycle, the range from the previous cycle is multiplied by this constant to give the new range, for translational and angular displacements.
If desired, the states can be sampled during a docking and output to a trajectory file. This file contains all the state variables required to define each sampled conformation, position and orientation of the ligand. The user can specify the range of cycles to be sampled. This allows the selection of the last few cycles when the docking will be nearing the final docked conformation, or the selection of the whole run.