<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type" />
<title></title>
</head>
<body>
<p style="margin: 0px;">Hi all,</p>
<br />
I have been running FEP simulations for over 6 months with little success. Our group is new to MD and FEP, and I and Ommi have been put in charge of figuring it all out. This is what we have been doing up to date:<br />
<br />
1. FEP of methane solvation (worked wonderfully)<br />
2. FEP of Riffampicin in RNAP (No idea due to little info on binding data)<br />
3. FEP of Biotin/Iminobiotin/Thiobiotin in avidin/streptavidin (great data on binding info but bad FEP results)<br />
4. Preliminary FEP on the statins due to good binding data and crystal structures (no good data yet)<br />
<br />
We have been running a 21 window FEP/BAR approach evenly spaced between lambda 0.00 and 1.00. (increments of 0.05) And then using the trapazoid rule and now the Bennett Acceptance Ratio to get the binding energy. Our system consists of the ligand in the protein at lambda = 0 with vdw-q and goes to lambda = 1 with none. We use the AMBER03 FF and TIP3P for the protein and water, and use gaussian to generate geometries and charges for our ligands. We then energy minimize with the steep function and then run a -DPOSRES FEP/MD run for 10ps to relax the waters and allow them to diffuse into the protein. We then run a FEP/MD with no defined values for various lengths, from 10ps to 5ns. We find that the longer we go the more precise our answers get but we never get any accuracy. If I can include any other data or mdp files let me know, I am stumped at what is needed to be done. <br />
<br />
Are we using enough windows?<br />
<br />
Is BAR the correct way of calculating or binding energies?<br />
<br />
Do we need to use a different FF?<br />
<br />
Do we need to have a longer diffusion POSRES MD?<br />
<br />
How long of a FEP/MD is long enough? (I know system size will have a huge influence on this answer)<br />
<br />
How precise is enough? (standard deviations of less than 2 kJ/mol in BAR?)<br />
<br />
Currently we are running long (2.5ns) relaxation MDs with no FEP and then running FEP on them for various times, hoping this will decrease any errors we are seeing.<br />
<br />
Thank you,<br />
<p style="font-family: monospace; white-space: nowrap; margin: 5px 0px 5px 0px;">TJ Mustard<br />
Email: mustardt@onid.orst.edu</p>
</body>
</html>