[AMBER] problems with relative TI calculations

From: <hannes.loeffler.stfc.ac.uk>
Date: Thu, 18 Jun 2015 09:35:42 +0000


I have run into some problems with relative TI calculations. I am interested in testing the relative hydration free energy of a set of small organic molecules. What I find is that the default 1-step protocol (simultaneous LJ and electrostatic softcores) in sander and pmemd does not work in some cases. One of them is neopentane > methane. The methane is mapped such that its carbon overlaps with the central C of neopentane. The 12 hydrogens of neopentane are defined as softcore atoms.

The 1-step protocol suggest a relative free energy of about 7 kcal/mol which is clearly too much as experiment suggests a value around 0 kcal/mol. To further test this I have run "absolute" TI simulations, a 3-step (complete decharging, vdw+bondend and recharging transformation; the latter is zero for methane in both solution and vacuum) and a 2-step protocol (the intermediate state is a neopentane with methane charges, the softcores have zero end-charges; the second step is a vdw+bonded transformation of this intermediate to the final state). All of these give me a free energy close to 0 kcal/mol so quite different from the 1-step protocol. An alternative mapping (methane C to terminal C in neopentane) gives me 1.2 kcal/mol in the 1-step protocol and kcal/mol in the 3-step protocol,.

I have also noticed in the neopentane > methane mutation that the final C-H distance of the methane end state is 1.22 Angstrom instead of the expected 1.09 A. This is regardless of chosen protocol. So I have also tested this with tert-butane and propane (methane again mapped to central carbon) to find end-point distances of 1.18 A and 1.14 A in vacuo.

I do not know what is going on here but some things clearly don't match up.

Many thanks,
AMBER mailing list
Received on Thu Jun 18 2015 - 03:00:03 PDT
Custom Search