[gmx-users] Pulling ligand out of protein for umbrella sampling

[Christopher Neale]

Dear Suhaib:

The topic you are asking about is imense. I suggest starting with some reviews. I haven’t done this type of thing in a few years, so I am not up on the newest developments, but there have been some very nice work by Mobley, Chodera, and Shirts (mostly together, but possibly separately also I think).

The issue that I mentioned in the post you cited was simply one of many considerations. The issue there was to try to reduce sampling less relevant possible 3D locations for the center of mass of the ligand with respect to the binding pocket (basically trying to pull it out through a cylinder rather than a cone, or even worse shapes). 

I think if I had to do this now I’d use some of the confine and releaase methods introduced by the authors I listed above (they developed it initially for alchemical decoupling, but it could also be useful in US), and I’d combine that with some type of or linear definition of the spatial unbinding order parameter.

I suggest that you get your feet wet with some simple systems with known answers before moving to your actual problem of interest.

Some other notes:
- there is no “easy” way to do this. In fact, the methods that are better (because they reduce convergence time) are much harder than the simplest approaches that will almost certainly five you a garbage answer in all but the most trivial cases.
- if you want more specific help, then you will get it more likely if you come to the list with a specific issue rather than a broad request for information on how you should proceed (that is unless anybody can simply link you to a set of tutorials that walks you through simple-to-hard systems and if such a tutuorial set exists then I’d also be keen to know about it!)

Chris.

— original message —

Add Reply
Suhaib Shekfeh 12 days ago
ReplyPermalinkRaw Message Dear GMX-users

In an old post in this list about umbrella sampling in gromacs and the pull
code, (
http://gromacs.org_gmx-users.maillist.sys.kth.narkive.com/w3NmJ69y/md-pull-code-in-umbrella-sampling
)
Christopher Neale gave an interesting comment about the case of pulling a
ligand out of a protein, especially when some of first contacts between the
two pull groups can cause improper sampling as they need other orthogonal
X/Y degrees of freedom

his suggested solution was as follows: "Personally, I would solve this by
adding a second pull group with a flat-bottom potential that acts only on X
and Y to keep the drug within a cylinder that extends along Z away from
your protein. However, you would need to modify gromacs source code for
this (I have some posts on the uses lists over the years about how to do
this). An alternative solution is to define an order parameter that acts in
all 3 dimensions, although you may need to compute the free energy of releasing
this restraint (a) in the binding pocket and (b) at large separations where
your PMF flattens out to zero."

I would like to ask Christopher or other expert users here to explain more
about these solutions, especially which one of them is easier?
--
Suhaib Shekfeh
Department of Pharmaceutical Chemistry
Gazi University Eczailik Fakültesi (Faculty of Pharmacy)