All,

Thank you for your help.
I have run the outlined computations using the suggested parameter substitution for two molecules of interest (1,6-Ac2-alpha/beta-D-glucose) and found the average absolute error to be 0.85 kcal/mol and 1.53 kcal/mol (9.6% and 19.9% relative to maximum barrier height respectively). I can share the exact results if there is any interest in this. The main deviation lays in the region between 0-120°, where MM energies overestimate for alpha, while underestimating for beta - making further parameterization tricky, as I understand it.

Comparing the results for the O5-C1-O1-C1A torsion above to the rotational energy profile for the C5-C6-O6-C1A torsion in 6-Ac-b-D-Glc (using the existing GLYCAM parameters), where I found 0.92 kcal/mol and 6.8% respectively, I think the parameter substitution for 1-Ac may not be ideal, but might be sufficient to use non the less?

 -Sven

On Wed, Sep 21, 2016 at 10:07 AM, Rob Woods <[log in to unmask]> wrote:
Hi Sven,
1) Do not include water (explicit or implicit) in the MM calc. You are trying to match vacuum QM data. Later in MD you will include water. 

2) I would set R2=R3=angle. 

3) Monitor the RMSD between the minimized MM and QM structures at each angle and make sure they are essentially the same.

Good luck,
Rob

Excuse the brevity, sent from iPhone

On Sep 21, 2016, at 12:28 PM, Sven Hackbusch <[log in to unmask]> wrote:

Thank you to everyone that has chimed in with your help!

I do have some follow-up questions about the MM energy minimization with the targeted torsion restrained:  

1. I assume that the minimizations are run in explicit solvent (TIP3P model) - is this correct?
2. With what values do you restrain the torsion angle in your minimizations? In my first attempt, I kept to the AMBER tutorial 4 by setting r1 = angle - 2°, r2 = angle - 1°, r3 = angle + 1°, r4 = angle + 2°  and rk2 = rk3 = 32.0, however this seems arbitrary to me and I did not find anything given in the GLYCAM06 publication.

Thank you!
  -Sven

On Sun, Sep 18, 2016 at 3:43 PM, Rob Woods <[log in to unmask]> wrote:
Hi Sven,
To answer your specific question, yes, we normally would develop them using small representative training molecules. I would not expect there to be a torsion term required for the H2. My rule is that if it is small, it's preferable to use zero. But that is system dependent. 

Please remember to fit the charges with the aliphatic Hs constrained to have zero net charge. 

Good luck!
Rob

Excuse the brevity, sent from iPhone

On Sep 18, 2016, at 11:50 AM, Xiaocong Wang <[log in to unmask]> wrote:

Hi Sven,


To develop torsion terms for missing terms in GLYCAM06, you will need to generate QM energy profile for that specific torsion.  To do that, optimize the molecule with that torsion restrained in QM (HF/6-31G*), then single point energy calculation with the optimized structure (B3LYP/6-31G*).  We usually vary the torsion angle every 30 degrees.


Once we have QM optimized structures (targeted torsion restrained), perform MM energy minimization with targeted torsion restrained as well.  For parameters you use in MM energy minimization, set V terms for that specific torsion to 0.  You will have the MM energies without rotational energy corrections ("torsion energies").  The differences between QM and MM (without rotational energy corrections) energies are what "torsion energies" are supposed to be.  Then, develop V terms to make sure rotational energies (sum from cosine functions) match the previous energy differences.


It is not complicated, but it will take some time.  The logic behind torsion terms development is that torsion terms (rotational energies) are corrections in MM energy calculations.  Therefore, if the similar terms are fairly small, it means you could either copy from similar terms or set the V terms to 0.  Besides finding similar terms in GLYCAM06, you can also search similar ones in parm and gaff, which are available in AMBER.


Best,


Xiaocong Wang
Complex Carbohydrate Research Center
The University of Georgia
315 Riverbend Road,
Athens, GA, 30602

From: Users of GLYCAM & GLYCAM-Web <[log in to unmask]> on behalf of Sven Hackbusch <[log in to unmask]>
Sent: Friday, September 16, 2016 3:09:18 PM
To: [log in to unmask]
Subject: Re: GLYCAM parameters for 1-Acetylated sugars
 
Hello Yohanna,

Thank you for your offer to help!

I think I will first try to use the approximation you recommend and compare the results to QM computations to verify it is a good enough approximation.
If I understand the Glycam06 paper correctly, I will need to generate a rotational energy curve for the rotation about the Cg-Os bond for my molecule at the B3LYP/6-31++G(2d,2p)//HF/6-31G* level of theory and then to compare this curve to the GLYCAM06 result.
--Can you explain how I generate the rotational energy curve from the GLYCAM06 parameters?

If I were to have to develop my own parameters, would it be correct to use two simple compounds (methyl methoxyacetate and tetrahydro-2H-pyran-2-yl acetate) as a training set to generate the parameters?
Here, I am not completely clear on how the fitting of the parameter terms was performed (in terms of the math involved) and how the rotational energy curve was divided into the respective atomic sequences (as opposed to using just a single atomic sequence in GLYCAM93).

Thank you again!
  -Sven

On Fri, Sep 16, 2016 at 7:17 AM, Yohanna White <[log in to unmask]> wrote:
Hello Sven, 

You have several options for this case. You can either make approximations and make a guess on the torsion terms, develop your own parameters, or we can develop the parameters for you and guide you through it (but this will take us about 6 weeks to get to). 

So if you'd like to make some approximations, you could set  the Os-Cg-Os-C to the same term as a Cg-Cg-Os-C, assuming that if this were a hexane, its torsion term would be similar. And you can set the H2-Cg-Os-C to 0 so it won't have a torsion term. 

Let us know if which plan works best for you. 

Thanks,
Yohanna 

On Tue, Sep 13, 2016 at 9:30 PM, Sven Hackbusch <[log in to unmask]> wrote:
Dear GLYCAM developers and users,

I am interested in simulating a sugar that is acetylated at the 1 position, specifically 1,6-Ac2-alpha-D-Glc, with the GLYCAM forcefield in Amber.
It would appear that there exist no partial charges for this compound (the online builder did not let me place an acetyl derivative at position 1), so I am using the general protocol for deriving the partial charges.
However, in generating the prmtop file using xleap I am encountering the following errors:

 ** No torsion terms for Os-Cg-Os-C
 ** No torsion terms for H2-Cg-Os-C

Presumably, GLYCAM is missing the dihedral angle parameters for the O5-C1-O1-C(acetyl) and H1-C1-O1-C(acetyl) dihedral angles.

Can anyone advise a way to overcome this roadblock?
Thank you very much for your help!

  -Sven


--
Sven Hackbusch
PhD Candidate
Department of Chemistry
University of the Pacific




--
Sven Hackbusch
PhD Candidate
Department of Chemistry
University of the Pacific




--
Sven Hackbusch
PhD Candidate
Department of Chemistry
University of the Pacific




--
Sven Hackbusch
PhD Candidate
Department of Chemistry
University of the Pacific