Re: question

From: Jérôme Hénin (jerome.henin_at_ibpc.fr)
Date: Mon Mar 31 2014 - 12:26:01 CDT

Hi Jay,

Indeed, you probably need to do more homework.

The main problem you'll run into is transferability. Your QM charges are good for one conformation, but nothing tells you they'll remain valid once you do MD and the local conformation changes. What happens when a hydrogen bond forms or breaks? Point charges are not observables but empirical parameters within an effective potential energy function. They have to be consistent with the rest of the force field, otherwise the effective potential is just effectively wrong - except if you go full QM for the whole interaction potential. And even then, there is nothing in ab initio that is intrinsically superior to empirical. Some empirical water models do a much better job of describing liquid water than many ab initio models.

To be perfectly honest, the transferability of the point charges from the force fields is not that great, but they are typically a compromise of some kind, and they are designed with the problem of transferability in mind.

If you still want to use your charges, you can just replace the charge values from the force field in the PSF file with your own, either with a custom program or a VMD script (changing the charge property of an atom selection). That's unless you use the AMBER file formats, which I don't know enough to help you with.

Cheers,
Jerome

----- Original Message -----
> I've gone through the first VMD and NAMD tutorials and have read
>
> around in various articles but still am puzzled with this question:
>
>
> I'm doing quantum mechanical calculations with our own local software
>
> and have partial charge results on a whole protein. It seems MD software
>
> only uses partial charges from model peptides, etc. and not the result from
>
> the full moelcule, so my question is:
>
>
> How do I use these partial charges in NAMD rather than the partial charges
>
> that are begun with from model compounds?
>
>
> I understand there's a lot of parameterization to emprical values but I have
>
> an ab-initio result. I also could produce the electronic potential and have a
>
> wave function as well.
>
>
> Really I want to be able to link our results into MD simulations (of course)
>
> like NAMD or perhaps Amber which uses the RESP method.
>
>
> Maybe I need to do more homework but everyone seems to avoid addressing
>
> this question.
>

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:16 CST