Re: TI alchemical transformation of a ring atom: small partial charge difference affects strongly dG

From: Mateusz Bieniek (bieniekmat_at_gmail.com)
Date: Tue Jul 21 2020 - 07:33:17 CDT

Thanks for your reply Peter. It seems the problem where the issue
originated actually was related to sampling. I had really poor results from
my protein simulations which I cannot reproduce, and these results threw
off my ddG. Please close this issue. The answers below are for the sake of
completeness.

file is a mol2 file without any specification of incoming/outgoing atoms

C7 is morphed into C59, and CL8 is morphed into BR1 in the .mol2 file.

 Where do you derive your expectation from?

I expect that the internal atom to the ring, having its charge change only
by less than 0.1e, should not change my dG by 3 kcal/mol. The atom does not
appear to interact with 1) the environment particularly match, 2) the rest
of the molecule, to justify such a large magnitude. Now with the new
results I can see that I was wrong in this case and this was not an issue
at all.

Why isn't just the single atom Cl->Br perturbation the right way to go?

The reason why Cl-Br are not the only atoms being mutated is due to the
automated protocol which uses a cutoff value for charges to decide whether
other atoms need to be mutated. Regardless of the cutoff value, this case
can show up in other places.

Do the parameters for anything else in the ring have to change?

The rest of the ring is exactly the same.

Could you also check and confirm that the net charge of the
> incoming/outgoing groups is equal in each case?

In the attached example/test the disappearing part has q=-0.1244, whereas
the appearing q=-0.1992. However, in the original case they were equivalent
(-0.16185 and -0.16175).

Thanks for your help,
Mateusz

On Mon, 20 Jul 2020 at 02:54, Peter Freddolino <petefred_at_umich.edu> wrote:

> Dear Mateusz,
> Unfortunately because your posted file is a mol2 file without any
> specification of incoming/outgoing atoms (and your imgur link leads to a
> blank page), it is difficult to figure out exactly what is going on here.
> Could you begin by giving a clearer description of the various ways that
> you've attempted to run this, and the results that you got? Where do you
> derive your expectation from? Why isn't just the single atom Cl->Br
> perturbation the right way to go? Do the parameters for anything else in
> the ring have to change? Dealing properly with the bond and angle terms in
> a ring system during decoupling is tricky business. Could you also check
> and confirm that the net charge of the incoming/outgoing groups is equal in
> each case?
> Best,
> Peter
>
> On Sat, Jul 18, 2020 at 11:39 AM Mateusz Bieniek <bieniekmat_at_gmail.com>
> wrote:
>
>> Hi everyone,
>>
>> I am transforming a ring into the same ring by mutating only one atom.
>> That's the C7 ring here https://imgur.com/1vSkZ8k
>>
>> Mutating just the sticking out heavy atom Cl/Br works fine and so far is
>> the only way for me to obtain correct energies.
>>
>> One of the atoms has a slightly different charge, and I need to mutate it
>> (it is an automated protocol). The small charge difference is 0.12e.
>> However, that mutation leads to very large dG in the electrostatic
>> component (3.5 kcal/mol, reproducible). When I match the charge on the
>> incoming and outgoing ring atom, the problem disappears. If I reduce the
>> charge, the dG electrostatic integral decreases substantially.
>>
>> I have so far used only the dual topology approach in NAMD. However, I am
>> surprised that such a small change in the charge leads to dG blowing up. I
>> expect the ring to be solid because I do not degrade the bonds during the
>> transformation. Therefore, the atom should remain in the same position.
>> Furthermore, this atom should have the same position with respect to the
>> ring.
>>
>> I added a rather naive test where I set scaling14 to 0.2 value. This has
>> dramatically reduced the 3.5 kcal/mol to 1.5 kcal/mol. Although it is
>> tricky to conclude anything. Nevertheless, the ring atom C7 transformation
>> introduces a huge change in the electrostatic component, but I cannot see
>> how C7 could interact with anything, and how such small changes in the
>> partial charge could change dG so much.
>>
>> The combined molecule:
>> https://pastebin.com/uZQ6CCg1
>>
>> Many thanks,
>> Mateusz Bieniek
>>
>

This archive was generated by hypermail 2.1.6 : Fri Dec 31 2021 - 23:17:09 CST