Bonjour,

I have a protein in PDB format that I have used successfully, to make a model (using the command "a.very_fast()"). However, whenever optimization is carried out (even by the default optimization in "model-full.py") some of the residues convert to their D-isomeric form.

To avoid this I have tried the following:
  1. Making a series of models rather than just one to avoid problems caused by randomization.
  2. Turning randomization off using the "deviation=None" command in the "allhmodel" parameters.  
  3. Using a number of cycles of optimization with a small number of iterations rather than one optimization step with many iterations.
  4. Finally I tried using 2 files from the RCSB protein data bank, with post numeration (ie atoms are labelled HD1 rather than 1HD). This was to check if the error was from my PDB file.

This did not fix the problem.

Further information regarding the problem is as follows:
  1. The conversions only take place after VTFM optimization is conducted in the "model-full.py" file. Using "a.very_fast" does not cause conversions.
  2. The conversions occur in a random fashion.
  3. My original PDB file contains 10 identical chains, one on top of another. Conversions that occur in one chain do not occur at the same location or residue in other chains.
  4. Residue type does not affect the chances of a residue becoming converted.
  5. Finally, I get these warnings in the ".log" file:

Thank you in advance for your help, and Happy modelling,
Abdullah Ahmed

PS I have attached two PDB files as an example of the problem. If the  two files are super-imposed on each other (using pymol, for example), you will find that residue number 9, GLU has switched to a D-isomer. Is is evident by the fact that Hydrogen atom in this residue is rotated 180° with respect to the other file. 

Hotmail: Trusted email with Microsoft’s powerful SPAM protection. Sign up now.