AMBER Archive (2003)

Subject: SANDER vs GIBBS equilibration

From: Sophia Kondratova (x357l_at_unb.ca)
Date: Wed Mar 26 2003 - 07:44:39 CST

I have noticed something peculiar while trying to perform an FEP
simulation of phenol --> benzene in gas phase using the GIBBS module
of AMBER7.

Before doing the FEP simulation I equilibrated the system using MD for at
least 35 ps with a time step of 1 fs ( I have done longer equilibrations as
well for up to 4035 ps). The FEP was performed in gas phase using 21
windows of 50 ps equilibration and 50 ps production steps ( the FEP
direction is lambda1-->lambda0).

Using a perl script to extract etotal, ekinetic, epotential ... from the gibbs
output file I have graphed the potential energy vs time for each of the 21
lambda runs of the FEP. I have put this graph on a server
http://biocomp.chem.unb.ca:8080/GD/FEP-graph.html

As can be seen on the graph, it took at least 170 ps of the lambda1 gibbs
run to equilibrate before the potential energy leveled off for the rest of the
21 runs.

Should there be such a transition when a well equilibrated structure from
SANDER is inputted into GIBBS for an FEP simulation if all the
parameters between SANDER and GIBBS are the same ?