OEDocking TK 1.2.3¶
The following improvements have been made to OEPosit:
- Graph comparison is used to identify when default OEOmega sampling may not produce appropriate conformers. If this is detected, the ShapeFit methodology is used to further sample conformation space.
- The initial overlays for ShapeFit are now chosen with the best TanimotoCombo overlay to the bound ligand prior to optimization.
- The choice of fitting routine (ShapeFit, Hybrid, or FRED) now take into account both the TanimotoCombo overlay and the graph similarity of the molecule. The graph similarity helps during cases where OMEGA undersamples and by sampling more conformations, better overlays can be generated.
- When generating conformer ensembles, the maximum number of conformers is set to 100 times the number of rotatable bonds with a minimum number of 200 conformers. This allows better sampling of floppy conformations.
- The maximum MMFF94s induced strain is increased to 20 kCal for molecules with more than 12 rotatable bonds.
- When FRED or HYBRID can not fit a molecule into the receptor, ShapeFit is automatically chosen regardless of the expected probability.
- If the input conformations include 3D coordinates, they are retained even when generating conformations.
Major bug fixes¶
- OEDocking::OEPosit::DockMultiConformerMolecule would previously only generate one pose. Now up to numPoses poses will be generated.
- SDData is retained from the input molecule. In the case of an input conformer ensemble, the SDData is chosen from the conformer with the closest RMSD to the final optimized pose.
Minor bug fixes¶
- A GOOD pose has been set to a probability of >= 0.5 not > 0.5 to be the same as specified by the documentation.
- Removed unbounded stack allocations.
- Removed annotation of optimization steps when writing to .oeb format.
- Rare crash in ShapeFit fixed for C++ and Python toolkits on OSX 10.8 and 10.9.