AMBER Archive (2008)

Subject: Re: AMBER: how to derive torsional angle parameters in gaff

From: Balazs JOJART (
Date: Thu Jun 26 2008 - 03:53:40 CDT

Dear Dr. Wang, and Develpoers,
Yesterday I consulted with my co-workers, and we have some questions
about the dihedral fitting procedure.
We see, that you used gaussian98 programpackage, wich can be resulted in
a minimal difference, compared to g03 results.
Quting from your answer to Dr. Case 'my one is total energies and
Balazs' is the total energy minus Ebond minus Eangle minus
Enonbonded.'. Is that mean, that in GAFF torsion development the QM
torsional profile was used instead of the QM-MM(non-bonded)? If this is
true, than (as far as I understand the development) the obtained
parameters are not good based on the AMBER8 manual: 'Before Fitting the
torsional parameters, we must generate the energy profile for the
molecular mechanical nonbonded potential as was done for the quantum
potential, subtract this curve from the quantum curve, and fit the
torsional potential to the difference potential', which means that we
have to fit to the difference curve, not to the initial QM curve.
You wrote 'different protocols in scanning calculations'. In the GAFF
papaer the following strategy was described:
'the whole search range is 180° (except in our case where the range was
-180 - 180) with a step size of 30°; for each point, minimization was
performed at MP2/6-31G* level then followed by a single calculation at
the MP4/6-311G(d,p) level. I discussed this with the ab initio guys, and
we concluded, that this is a typical relax scan with g03. we performed
the same protocol in our calculations. Or we have to perform rigid scan
for the molecule? But in this case we have to subtract the
MM(non-bonded) curve from the QM curve, as well.
You mentioned also the zero-point vibrational energy, which was not
inculded in our calculations. This energy is due to the kinetic energies
of the nuclei of the molecules. But we have to calculate the potential
energy of the system at the MP4 level.
I'm a little bit confused within this field.
Would you be so kind to clarify these points in more detail?
Thank you very much for the help and discussion.
Balazs Jojart

Junmei Wang wrote:
> The following is the emails between Balazs Jojart, David Case and me.
> Maybe you are interested in this topic.
> Junmei
> Hi Dave,
> Now I understand why the two figures are so different: my one is total
> energies and Balazs' is the total energy minus Ebond minus Eangle
> minus Enonbonded. As a matter of fact, X-c-c-X has a V2 term in gaff,
> and this is consistent to Balazs's figure, which has minimum at -180,
> 180 and 0.
> The attached pdf files have rotation profiles for GAFF, QM (total
> energy), MM2 and MMFF. They are all quite similar. GAFF has similar
> rotational barrier to QM, while MM2 and MMFF have much big barriers.
> The discrepancy between Balazs' QM profile to Figure 3 in the paper
> may be caused by (1) ZPE, (2) different protocols in scanning
> calculations. I am not sure if Gaussian 03 has improved 'scan' or not.
> In Gaussian 98, 'scan' first optimizes the structure and does single
> point calculations for every conformations. However, I don't think
> this is an ideal protocol to generate rotational profile. For Figure
> 3, I performed 'popt' with the involved torsional angle frozen.
> Best
> Junmei
> Hi Balazs,
> I am sorry for not replying your email earlier. I didn't re-run the QM
> calculations to verify your QM data, I just perform torsional angle
> scanning using MM2 and MMFF force fields. I got similar profiles as
> the one in Table 3 of GAFF paper. I will double check this issue when
> I upgrade gaff to gaff2 or uaff (universal amber force field). Thank
> you very much for your work on gaff development.
> Please find the attached pdf file for my rotational profile calculations.
> Best
> Junmei
> ----- Original Message ----
> From: Balazs Jojart <
> <>>
> To: <>
> Sent: Thursday, June 5, 2008 5:42:02 AM
> Subject: gaff parametrization
> Dear dr. Wang,
> I wroe a mail directly to dr. Case yesterday, and he suggested me to
> forward my message to you, because you prepared the torsional parameters
> for the general amber forcefield.
> You can read the mail as follows, and I attached the files, as well.
> Dear dr. Case,
> I write you diretly because you are the coresponding author of the
> article 'Development and Testing of a General Amber Force Field'.
> I started 3 or 4 weeks ago to learn the development philosophy of the
> amber forcefield, especially the torsional parameter fitting.
> I thought, the the paper mentioned above, is a good starting point,
> because it contains the torsional profile of acethyl-acetone
> (ch3-co-co-ch3).
> Unfortunatelly, I wasn't able to reproduce the data (figure 3.)
> I performed the following protocol:
> (1) the small molecule (MOL) was prepeared using gaussview
> (2) optimization was performed at the MP2/6-31G* level, anf frequency
> calculations were also conducted. I found only one imaginary freqency,
> but rotating 3 degree of a methyl rotor solved the problem.
> (3) relax scan was performed on MOL with the following keywords:
> MP2/6-31G* Opt(Z-matrix,maxcycle=100) nosym, and for the appropriate
> dihedral angle : D2 179.97084591 S 11 -30.0
> (4) I prepared the input files form the gaussian log file using the
> gaussview program.
> (5) MP4 single point calculations were performed with the following
> keywords: MP4/6-311G**.
> (6) the UMP4(SDTQ) energies were collected from the single point
> calculations. (you can find it in the attached excel (GAFF_xccx.xls)
> file in column D3:D14, the values were converted to kcalmol-1 --> column
> E3:E14).
> (7) for the lowest energy conformer optimization at the HF/6-31G* and
> MEP calculation were performed (you can find the prepared prepi file as
> attachment (MOL.prepi)).
> (8) molmech minimization using AMBER 9 (input file
> <>), and the
> appropriate dihedral angle was restrained during the minimization with a
> 100 force constant (RST).
> (9) the BOND, ANGLE, VDW, ELE, 1-4 VDW and 1-4 ELE were collected form
> the log file file (G:N).
> (10) in the Amber8 manual a read the following :'Before Fitting the
> torsional parameters, we must generate the energy profile for the
> molecular mechanical nonbonded potential as was done for the quantum
> potential, subtract this curve from the quantum curve, and ?t the
> torsional potential to the difference potential'
> (11) I calculated the MP4-MM(bond,angle,ele,vdw,1-4)[D18:D29] and
> MP4-MM(ele,vdw,1-4) [D34:D45].
> (12) the obtained curves are completly different compared to Figure 3 in
> the article.
> My qustetion are:
> Did I made any mistake during the calculation? I don't understand the
> parameter development strategy in amber?
> For the X-c-c-X torsional term th V2=1.2 kcal/mol, multiplicity=4, phase
> angle=180. It is not clear for me how came these values from Figure 3.
> 'PHASE = 180 degrees if an energy minimum is at 0 degrees.' 8form amber8
> manual. As we can see on Figure 3 there is a maximum at 0 degree. The V
> paranmeter is not the one-half of the barrier? n=4, which means the
> there is 4 minimums when during the rotation from 0 to 360 degree.(AR
> LEach book).
> Etors can be calculated with the following equation:
> summaVn/2(1+cos(n×phi-phase)). Am i right? Consider only the first trhee
> term:
> V1/2(1+(cos(1×phi-phase))+V2/2(1-(cos(2×phi-phase))+V3/2(1+(cos(3×phi-phase))
> (as I read the article of Jorgensen: Development and Testing of the OPLS
> All-Atom Force Field ... (J. Am. Chem. Soc. 1996, 118, 11225-11236).
> Then where can we set up the multiplicity parameter? The 'n' parameters
> are not the same in the equation? Beacuese in the Gaff paper the 'n' is
> described as the multiplicity, but in the Jorgensen paper n is equalt to
> n in the Vn term.
> Can you give me some notes where I made a mistake during the calculation
> and the protocol?
> Your help will be very appreciated!!!
> Thank you for your help in advance!
> Balazs Jojart Ph.D.
> University Of Szeged
> Department of Chemical Informatics

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" (in the *body* of the email)