[gmx-users] About "Invalid order for directive atomtypes" error

[Hans Martin Senn]

Hi Xavier

On 16 Jul 2008, at 15:51, Xavier Periole wrote:
> On Wed, 16 Jul 2008 15:04:17 +0100
> Hans Martin Senn <senn at chem.gla.ac.uk> wrote:
>> Dear all
>> This is a topic that has come up a couple of times here, see, e.g., http://www.gromacs.org/pipermail/gmx-users/2008-January/032114.html 
>>  .
>> However, despite the recommendations given in that previous post,  
>> I  think the problem is exactly that the information in the manual  
>> is not sufficiently accurate or clear in this case.
>> Specifically: I have a protein with two bound ligands, ZTRP and FAD.
>> - For the protein I use ffamber03. pdb2gmx created two itp files  
>> for  the two chains.
>> - For ZTRP, I have built an itp file using AMBER atom types (as   
>> defined in ffamber03.atp or ffamber03nb.itp).
>> - For FAD, I have an itp file generated using antechamber and  
>> amb2gmx.  It contains an [ atomtypes ] section defining the GAFF  
>> atom types.
>> This was my first go at assembling it all in the top file:
>> >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>> ; Include forcefield parameters
>> #include "ffamber03.itp"
>> ; Include chain topologies
>> #include "topol_A.itp"
>> #include "topol_B.itp"
>> #include "topol_FAD.itp"
>> #include "topol_ZTRP.itp"
>> ; Include water topology
>> #include "ffamber_tip3p.itp"
>> ; Include generic topology for ions
>> #include "ions.itp"
>> [ system ]
>> ; Name
>> Full system in water
>> [ molecules ]
>> ; Compound        #mols
>> Protein_A           1
>> Protein_B           1
>> ZTRP                  1
>> FAD                   1
>> SOL             20893
>> >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>> However, grompp complained about "Invalid order for directive   
>> atomtypes" when including topol_FAD.itp.
> It might be that the same protein name is in two of your different
> topology files.
> The problem is certainly somewhere in those topologies.

By "protein name", do you mean the [ moleculetype ] section? If so,  
the answer is no. Each of the molecule itp files specifies exactly one  
molecule type (protein chain or ligand); the names are unique.
Of course it is fair to say that the problem is in the topologies, as  
demonstrated by my solution to the problem... The point is that it is  
apparently not permitted to have [ atomtypes ] (in topol_FAD.itp)  
after having read other molecule definitions (in topol_{A,B}.itp), in  
contrast to what I think the manual suggests.

Hans



>> Following a hint from the mailing list, I took the [ atomtypes ]   
>> section out of topol_FAD.itp and put it into a separate itp file  
>> "GAFF- types.itp", which is included before the other molecule itp  
>> files. The  final top file then looked like this:
>> >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>> ; Include forcefield parameters
>> #include "ffamber03.itp"
>> ; Define GAFF atomtypes here (not in molecule itp file) to avoid   
>> 'Invalid order' error
>> #include "GAFF-types.itp"
>> ; Include chain topologies
>> #include "topol_A.itp"
>> #include "topol_B.itp"
>> #include "topol_FAD.itp"
>> #include "topol_ZTRP.itp"
>> ; Include water topology
>> #include "ffamber_tip3p.itp"
>> ; Include generic topology for ions
>> #include "ions.itp"
>> [ system ]
>> ; Name
>> Full system in water
>> [ molecules ]
>> ; Compound        #mols
>> Protein_A           1
>> Protein_B           1
>> ZTRP                  1
>> FAD                   1
>> SOL             20893
>> >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>> This solved the problem. However, and this is my main point here,  
>> the manual for GROMACS 3.3, p. 98 says:
>> "starting with GROMACS version 3.1.3 all directives at the  
>> parameter  level can be used multiple times and there are no  
>> restrictions on the  order..."
>> In my reading, this means that [ atomtypes ] can be defined  
>> anywhere  (as long as it is before they are referred to for the  
>> first time),  which is clearly at variance with the observed  
>> behaviour. Am I  misinterpreting it, or is the manual not precise  
>> enough here?
>> A final note: Since simulating a protein with bound ligands is an   
>> extremely common case and since this issue about "Invalid order"  
>> seems  to come up regularly, it would be very helpful if a  
>> corresponding  HOWTO was available in the manual/FAQ/wiki to settle  
>> it once and for  all.
>> Best wishes
>> Hans


.......................................................
Dr. Hans Martin Senn
Lord Kelvin/Adam Smith Research Fellow

E-mail: senn at chem.gla.ac.uk
Direct line: +44 (0)141 330 6574
Fax: +44 (0)141 330 4888

Department of Chemistry
Joseph Black Building
University of Glasgow
Glasgow G12 8QQ
Scotland, UK