I want to explore the low energy states of an amino acid or a peptide by generating input geometries, defined by multiple dihedral angles of the backbone or the side chain. From these starting conformations I want to do geometry optimization to yield local energy minimum geometries.
The following approaches failed:
- Using the ModRedundant keyword in the Gaussian09 package to define new values for the dihedral angles. This crashes for larger molecules and major changes in the coordinates.
- Detecting the substructure of the molecule (e.g. the backbone) by SMILE / SMARTS pattern and using obrotate from the openbabel package to change the values. Works fine for a single dihedral or maybe two, but afterwards these changes the pattern of the substructure is not necessarily detected anymore and no further dihedral angle can be changed.
- Usage of a well chosen Z-Matrix: The great disadvantage is, that I don't know how to yield a Z-Matrix with the wanted dihedrals from a given geometry and therefore I can't start from a pre-optimized structure.
Are there any other attempts to this problem?