Re: [AMBER] MM-GBSA on large protein/nucleic acid systems

From: Carlos Simmerling via AMBER <amber.ambermd.org>
Date: Wed, 7 Oct 2026 15:40:55 -0400

I think for large systems it is typical to parallelize by splitting the
frames across many cpus/threads if possible. Since they were generated
in advance you don't have to do them serially like an MD run.
I would be very careful in this situation about the solvent model - it
should be known to work well on proteins, RNA and small molecules. That's a
challenge, and many of the GB models are specifically trained for very high
speed and giving forces. They definitely compromise the accuracy to achieve
that speed. In MMGBSA, you typically have far fewer structures to evaluate
than in an MD run, and you don't need forces. That points to using a much
better implicit solvent model, which unfortunately many people don't do.
For example, as the developer of the GBneck2 models (igb=8 in Amber) I
would *strongly discourage* anyone from using it to estimate ligand binding
affinities. It just wasn't trained for that, and isn't the right tool for
the job.
A good place to start (again my opinion) is the "GBNSR6" model - it's
slower and doesn't provide forces, but it has far fewer parameters and
should be fairly insensitive to the types of molecules you use. In
contrast, GBneck2 is only trained for proteins and nucleic acids - if you
give it a ligand with a halogen for example, it really won't know what to
do with it but make a guess. NSR6 should do much better with things that
don't look like biopolymers.

--
Carlos Simmerling, Ph.D. (he, him, his)
Marsha Laufer Endowed Professor of Physical and Quantitative Biology
Professor, Department of Chemistry
Associate Director, Laufer Center for Physical and Quantitative Biology
Room 119 Laufer Center, Stony Brook University, Stony Brook, NY 11794-5115
http://www.simmerlinglab.org
On Wed, Oct 7, 2026 at 3:33 PM Chou, Ying via AMBER <amber.ambermd.org>
wrote:
> EXTERNAL EMAIL - This email was sent by a person from outside your
> organization. Exercise caution when clicking links, opening attachments or
> taking further action, before validating its authenticity.
> Hello,
>
> Is there a way to efficiently calculate free binding energies using
> MM-GBSA of a large protein/nucleic acid system? I’m talking specifically
> about a ribosome 60S unit (~200,000 atoms). We are interested in small
> molecule ligands binding at a specific site in the ribosome, and the ligand
> has interactions with both the RNA and a ribosomal protein. If using
> MM-GBSA is not feasible (because computationally limiting), what other
> methods are appropriate to calculate free energies for such system? Thanks!
>
> Get Outlook for Mac<
> https://protect.checkpoint.com/v2/r01/___https://aka.ms/GetOutlookForMac___.YzJ1OnN0b255YnJvb2s6YzpnOmNhNDRjYWQ3NjgzMTE3OGJmNTEwODczNWY0OTdlNjFmOjc6MGNiOTo2NzAxYTFlYjNjZDcxOTU0ZDg3MGNhZjAwODU1NjBkYTUzNDYwYTUxNmExN2E2NWI3YjdiMTcxYmM4NWViNWNlOnA6VDpG
> >
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
>
> https://protect.checkpoint.com/v2/r01/___http://lists.ambermd.org/mailman/listinfo/amber___.YzJ1OnN0b255YnJvb2s6YzpnOmNhNDRjYWQ3NjgzMTE3OGJmNTEwODczNWY0OTdlNjFmOjc6YjM0NjplY2NlMTEwNmQ4YTgxZTdhZTBhNDRjNzQwYWY4MjRmYzBmNmVlMzFjZTU4YTIyYzMzMmFjMTE3N2Y1MzdkNWEwOnA6VDpG
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber

Received on Wed Oct 07 2026 - 13:00:05 PDT