VMD-L Mailing List

From: Mayne, Christopher G (cmayne2_at_illinois.edu)
Date: Tue May 16 2017 - 14:07:20 CDT

Stêphane,

Apologies for missing this.

I have just now looked at it. My only concern would be that you have used multiplicities of n = 1, 2, 3 where ParamChem has suggested that n = 3 is potentially sufficient (although this may be where the error comes from). Have you tried to fit the dihedrals using ONLY an n=3 term? Looking back, this was my question last time as well.

Regards,
Christopher Mayne

> On May 13, 2017, at 6:30 AM, ABEL Stephane <Stephane.ABEL_at_cea.fr> wrote:
>
> Dear Christopher,
>
> Last march, I sent you a pdf where I describe the parametrization of dihedrals of Prop DAO with fftk (see attached and my initial message below). Have you had time to see it?
>
> Best
>
> Stéphane
>
> --------------------------------------------------------------------
> Stéphane Abel, Ph.D.
>
> CEA Saclay DSV/ISVFJ/SB2SM
> Institut de Biologie Intégrative de la Cellule (I2BC)
> Bat 528, Office 138C
> Gif-sur-Yvette, F-91191 FRANCE
> Phone (portable) : +33 6 49 37 70 60
> ________________________________________
> De : ABEL Stephane
> Envoyé : dimanche 5 mars 2017 11:46
> À : Mayne, Christopher G
> Objet : RE:vmd-l: Optimization of torsion angles with fftk, large force constant
>
> Hello Christopher,
>
> Sorry for the for the long delay in responding. After the message (see below) I have checked my results and did additional verification and calculations which are presented in the attached pdf.
>
> My goal is to optimize the hydrophilic part of the Lauryl dimethyl aminoxide (LDAO) (slide 1, top) to be compatible with CHARMM force field. For that I have first used ParaChem webserver to see if the current CHARMM/CGenff parameters can be used to model the LDAO without further optimizations. It suggests that only two dihedral angles (CCNO and CCCH) should be re-optimized (see the tables in the slide 2 and 3). All the other parameters (i.e. atomic charges, bond and angles) should be ok ( their penalty scores <10) see slide 3). To optimize these dihedrals, I have used fftk with the propylDAO as a model compound and obtain the following fittings (see slide 4). My question is :
>
> 1) Is my approach correct?
> 2) despite the fact that the RMSE is quite good (0.508) for the two dihedral angles, do the value make sense for you since the force constant seem very high for me (in particular for the first multiplicity).
> 3) Moreover I optimize the each dihedral separately I obtained RMSE values significantly lower (see slide 5). I think it is normal and expected
>
> CCNO RMSE : 0.150
> dihedral {CG321 CG324 NG3P0 OG312} 4.300 1 180.00
> dihedral {CG321 CG324 NG3P0 OG312} 0.590 2 180.00
> dihedral {CG321 CG324 NG3P0 OG312} 0.700 3 0.00
>
> CCCH RMSE : 0.272
> dihedral {CG331 CG321 CG324 HGP5} 1.500 1 0.00
> dihedral {CG331 CG321 CG324 HGP5} 0.598 2 180.00
> dihedral {CG331 CG321 CG324 HGP5} 0.658 3 180.00
>
> Thank you again for your help and comments
>
> Best
>
> Stéphane
>
>
> --------------------------------------------------------------------
> Stéphane Abel, Ph.D.
>
> CEA Saclay DSV/IbItec-S/SB2SM & CNRS UMR 9198
> Institut de Biologie Intégrative de la Cellule (I2BC)
> Bat 528, Office 138C
> Gif-sur-Yvette, F-91191 FRANCE
> Phone (portable) : +33 6 49 37 70 60
>
> ________________________________________
> De : Mayne, Christopher G [cmayne2_at_illinois.edu]
> Envoyé : mercredi 1 février 2017 21:29
> À : ABEL Stephane
> Cc : vmd-l_at_ks.uiuc.edu
> Objet : Re: vmd-l: Optimization of torsion angles with fftk, large force constant
>
> Stéphane,
>
> There’s not really enough information here to make any specific claims about your parameters. I will note that dihedrals described by multiplicities > 1 tend to have one dominant term while additional terms refine the shape.
>
> Did you try the fitting only using n=3, as suggested from the ParamChem output?
>
> Regards,
> Christopher Mayne
>
>> On Jan 26, 2017, at 3:37 AM, ABEL Stephane <Stephane.ABEL_at_cea.fr> wrote:
>>
>> Dear vmd/fftk users,
>>
>> I am currently optimizing some torsions of the propyl diaminoxude molecule (CH3-(CH2)-NO(CH2)2 with fftk (vmd 1.9.3) but I am not sure if my approach is correct. For instance for the CCNO torsion* I obtain the following values** that fit well the QM curve (FINAL RMSE 0.174) obtained from GAUSSIAN
>>
>> Suggestion from ParamChem (CG321 CG324 NG3P0 OG312 0.2600 3 penality 58.0)
>>
>> **dihedral {CG321 CG324 NG3P0 OG312} 4.200 1 180.00
>> dihedral {CG321 CG324 NG3P0 OG312} 0.591 2 180.00
>> dihedral {CG321 CG324 NG3P0 OG312} 0.722 3 0.00
>>
>> As you can see that force constant for the first periodicity is quite higher than the other values and force constants for all the dihedral in the cgenn.ff parameter listing. Is it possible or I missing something?
>>
>> Thank you in advance for your comments
>>
>> Stéphane
>>
>>
>
> <LDAO_Optimization_Mayne.pdf>