Dear all
I am running amber16 / amberTools17
I have been trying to run Tempered-Hamiltonian replica exchange for a test system, penta-alanine at 4 different temperatures : 290.00, 370.0 , 470.00 , 594.00.
I am scaling my system on a temperature ladder, and to allow an potential energy overlap between so few replicas, I am using an Hamiltonian scaling on the water molecules, 'to cool them down' (solute tempered method, REST1), I then obtain a flatter and lower energy distribution for 'hotter' replicas, where I only warm up the protein.
After some tuning of the Hamiltonian scaling I get an nice overlap of potential energies of the different replicas.
But when I try to run the replica exchange simulation, one exchange every 10 ps with NVT, I obtain 'infinity' values in the rem.log file, and all exchange attempts fail.
for example :
# exchange 366
1 2 290.00 -13616.13 -12621.30 Infinity Infinity F 0.00
2 1 370.00 -13460.20 -14521.17 -Infinity Infinity F 0.00
3 4 470.00 -13312.10 -12404.33 -Infinity Infinity F 0.00
4 3 594.00 -13134.49 -14095.51 -718.29 -Infinity F 0.00
My hypothesis is that this behavior comes from the rule to calculate probability
Pi→j=min{1,exp(−β1[H1(x2)−H1(x1)]−β2[H2(x1)−H2(x2)])
if the two ladders (Temperature, Hamiltonian ) are inverted (as in my case) this way of calculating probability becomes very unfavorable to the exchange.
This other example using the same Hamiltonian ladder but temperatures closer to each other (replicas 2-4 are 'cooled down' by the Hamiltonian), in this case even if the replica 4 has a lower energy than replica 3 the exchange won't take place...
Anything I would be missing ?
# exchange 6
1 2 290.00 -13529.49 -12972.55 Infinity Infinity F 0.00
2 1 330.00 -13457.53 -14035.47 -485.70 Infinity F 0.00
3 4 350.00 -13966.67 -13764.42 -440.27 485.57 F 0.00
4 3 370.00 -14243.34 -14452.75 -476.72 -Infinity F 0.00
Any help/suggestion is welcome !
Marie
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Feb 07 2019 - 03:30:02 PST
