[MMTK] Problem with Energy and Gradients Calculation

khinsen at cea.fr khinsen at cea.fr
Sun Nov 27 23:46:50 CET 2005


On 27.11.2005, at 14:18, vanitha at cs.wisc.edu wrote:

> I again get Nan for another protein 1FSS.pdb. I'm confident that  
> there are
> no missing atoms in this protein. I have applied the Conjugate  
> Gradient

There are, this is even pointed out in the PDB file:

REMARK   5 THE FOLLOWING RESIDUES HAVE MISSING ATOMS (M=MODEL  
NUMBER;   1FSS  79
REMARK   5 RES=RESIDUE NAME; C=CHAIN IDENTIFIER; SSEQ=SEQUENCE  
NUMBER;  1FSS  80
REMARK   5 I=INSERTION  
CODE):                                           1FSS  81
REMARK   5   M RES CSSEQI   
ATOMS                                        1FSS  82
REMARK   5     HIS A 486    CG   ND1  CE1  NE2   
CD2                     1FSS  83
REMARK   5     SER A 487     
OG                                          1FSS  84
REMARK   5     GLN A 488    CG   CD   OE1   
NE2                          1FSS  85
REMARK   5     GLU A 489    CG   CD   OE1   
OE2                          1FSS  86
REMARK   5     SER A 490     
OG                                          1FSS  87
REMARK   5     LYS A 491    CG   CD   CE    
NZ                           1FSS  88

In fact, if I try to use 1FSS.pdb in MMTK, I get an error message  
long before I can even try to calculate an energy:

 >>> universe.addObject(Protein('1FSS.pdb'))
Traceback (most recent call last):
   File "<stdin>", line 1, in ?
   File "/Users/hinsen/hao/mmtk2/MMTK/Proteins.py", line 654, in  
__init__
     items = conf.createPeptideChains(model)
   File "/Users/hinsen/hao/mmtk2/MMTK/PDB.py", line 172, in  
createPeptideChains
     chain.findHydrogenPositions()
   File "/Users/hinsen/hao/mmtk2/MMTK/ChemicalObjects.py", line 804,  
in findHydrogenPositions
     raise ValueError('position of ' + a.fullName() + \
ValueError: position of .His486.sidechain.C_delta_2 is undefined

> I need to quickly figure out if it is the protein or something else  
> that
> is going wrong. Alternatively, can one of you point me to a protein  
> for
> which you are able to get an energy number using the following  
> code? I can
> then test my code for that protein to see if the problem recurs.

I just tried it on 1BP2, the example that you first reported yielding  
NaN. For me it works fine, the results are:

-7661.3757547
83316.7191957

The individual energy terms are:

{'electrostatic/pair sum': -13785.195598355707, 'cosine dihedral  
angle': 2639.6511855676904, 'harmonic bond': 541.48249951851085,  
'electrostatic': -13785.195598355707, 'harmonic bond angle':  
1483.5000172367438, 'Lennard-Jones': 1459.1861413287252}

What happens when you calculate the individual terms? Are they all  
NaN, or just some of them?

Konrad.
--
------------------------------------------------------------------------ 
-------
Konrad Hinsen
Laboratoire Leon Brillouin (CEA-CNRS), CEA Saclay,
91191 Gif-sur-Yvette Cedex, France
Tel.: +33-1 69 08 79 25
Fax: +33-1 69 08 82 61
E-Mail: khinsen at cea.fr
------------------------------------------------------------------------ 
-------





More information about the mmtk mailing list