[AMBER] Fw: [josiah.bones.anu.edu.au: Calculation of ligand restraints in TI]

From: Josiah Bones <josiah.bones.anu.edu.au>
Date: Thu, 17 Jun 2021 22:08:33 +0000

Hi Taisung, thanks for your reply.

I'll give a very quick overview of what I'm trying to do, to try and get an idea if I am on the right track.

I'm attempting a single simulation where V0=ligand in bulk solution, and V1=ligand bound to protein. This is due to charges on the ligands. I intend to use the 6DOF restraints on the bound ligand, while only restraining the COM of the ligand in bulk solution to ensure it doesn't interact with the nearby protein, but not limit its internal degrees of freedom. I believe the COM restraints don't need to be accounted for in any special way in the thermodynamic cycle?

In this particular case, I believe you are saying I would still need an additional simulation to calculate the dG of turning on the Boresch restraints for the bound state, and then use the analytical equation for these Boresch restraints. I refer to the recent paper, regarding these restraints:

"Additionally, Boresch restraints have been implemented, which can be used in an automated fashion for ABFE simulations with many diverse ligands."


"The implementation in AMBER20 also permits these restraints to be included in the overall alchemical transformation such that the component of the free energy arising from the restraints in the bound state can be computed in the same way as other force field terms."

I saw the option gti_output in the Amber20 manual, and wondered if this provided the contribution of the restraints within the single calculation. If this is not the case, would you have any recommendations for what settings I need to include in the different steps in order to capture the thermodynamic cycle correctly, and to automate the use of the Boresch restraints as detailed in the recent paper? I am comfortable with the overall running of the primary simulation (capturing dG of binding in one step) but not with the additional method required for capturing the Boresch restraints. Any help would be greatly appreciated.

Thanks again,


From: accuratefreeenergy.gmail.com <accuratefreeenergy.gmail.com>
Sent: Thursday, 17 June 2021 10:22 PM
To: amber.ambermd.org <amber.ambermd.org>; Josiah Bones <josiah.bones.anu.edu.au>
Cc: 'David A Case' <david.case.rutgers.edu>; abir.ganguly.rutgers.edu <abir.ganguly.rutgers.edu>; 'Darrin York' <york.chem.rutgers.edu>
Subject: RE: [josiah.bones.anu.edu.au: [AMBER] Calculation of ligand restraints in TI]

Hi Josiah,

        In the current AMBER20 TI implementation, restraints involving atoms in the scmasks will be scaled with respect to lambda, i.e., if an atom disappears at lambda=1 state, all restraints involving this atom will become zero at lambda=1 state. This implementation was designed for relative binding free energy--so that Ligand A and Ligand B can keep independent sets of restraints when doing A-->B.

        This behavior, however, does not change what you need to do for absolute binding free energy. I believe that a proper 6-DOF-restraint will not have any effect on the free energy of a non-interacting ligand. You still need to calculate the free energy of turning on the restraints of the ligand to the protein

        You also can recover the pre-AMBER20 behavior through changing the restraint lambda-dependence by enabling lambda-scheduling of "TypeRestBA" (AMBER manual Section 23.1.7). Let me know if you want to do so.



Date: Thu, 17 Jun 2021 01:13:46 +0000
From: Josiah Bones <josiah.bones.anu.edu.au>
To: "amber.ambermd.org" <amber.ambermd.org>
Subject: [AMBER] Calculation of ligand restraints in TI

Hi there!

I am just reaching out for some help as to the behaviour of AMBER20 pmemd.cuda in how it deals with the 6DoF restraints commonly employed in absolute binding free energy calculations, where the ligandís movement is restrained in a binding pocket. These calculations typically utilise two main simulations (one to decouple ligand from protein, and another for decoupling the ligand in water). From my understanding, there is then the requirement of running an additional simulation to capture the free energy of turning on the restraints of the ligand to the protein, and finally an analytical calculation to remove them.

>From reading recent works examining the changes in AMBER20 to the TI capabilities, I am suspecting, but not certain, that the restraints are already calculated in the ligand+complex simulation, removing the need for an additional simulation. Is this assumption correct, and is anything needed to be done to ensure the correct behaviour? And does this contract the requirements of such an ABFE calculation in a single smoothstep process using only two simulations: complex+ligand(restrained), water+ligand(unrestrained), and then the analytical calculation to account for the restraints?

Thank you kindly for any assistance,

AMBER mailing list

----- End forwarded message -----

AMBER mailing list
Received on Thu Jun 17 2021 - 15:30:02 PDT
Custom Search