We believe we have finally found how to run sander with up to 100 atom types.
The required changes are:
- in rdparms.F90, line 108 : nttyp > 5050 (instead of 1830 previously)
- in parms.F90, lines 30-35: the dimension of the arrays cn1, cn2 should be
increased from 1830 to 5050
With those changes, the system seems to behave correctly.
Best Regards
Elise Duboué-Dijon
> Dear Amber developpers,
>
> We are trying to perform MD simulations using AMBER11 on a system
> containing a protein, a cofactor, a substrate, and solvent molecules. Our
> system contains 69 different atom types.
>
> Sander immediately stopped with an error because nttyp (defined as
> ntypes*(ntypes+1)/2 ) exceeds the limit of 1830. However, the maximum
> number of atom types (max_atom_types) is set to 100, which would lead to a
> maximum value for nttyp of 5050. So we changed the maximum value of nttyp
> to 5050=100*101/2 in the rdparm.f routine.
>
> With this patch, sander runs, but leads to huge 1,4-VDW energies (the
> prmtop file was generated with chamber from a charmm force field, and we
> checked that the procedure works fine on subsystems with fewer than 60
> atom types). Those 1,4 energy terms are 10000 times higher than the same
> terms calculated with NAMD. All other energies seem to be correct. We
> tried to minimize only parts of the system, and as soon as our subsystem
> contains less than 60 atoms everything is fine, but each time there are
> more than 60 atom types this 1,4-VDW energy gets enormous.
>
> Could you please tell us if this abnormal behaviour of the 1,4-VDW
> energies could actually be related to the number of atom types which is
> too large ? What else (in addition to the upper bound of nttyp in
> rdparm.f) should we change to be able to work with more than 60 atom types
> ?
>
> Thank you very much for your help.
>
> Best regards,
>
> Elise Duboué-Dijon,
> PhD Candidate
> Ecole Normale Supérieure, Paris, France
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Nov 27 2012 - 03:30:01 PST
