5.1 Starting positions for optimization

Since OEBestOverlay performs an optimization, the starting point of the optimization is important. By default, OEBestOverlay uses an inertial frame alignment method to decide on starting positions. The reference structure is aligned by its principal moments of inertia, then the fit object is aligned in 4 positions with the primary and secondary moments of inertia in both possible directions.

Thus, for any given optimization, there are 4 results returned, usually only one of which is useful. In order to deal with structures with symmetrical moments of inertia, OEBestOverlay may perform additional starting points. For a reference or fit where the 2 major moments of inertia are equal (to a user-defined percent, nominally 15%), 4 extra starting positions are generated. In the rare case of a molecule with all 3 moments of inertia essentially equal, 20 random starting translations and rotations are chosen as starting positions.

Alternatively, the user may desire to start the optimization from a single, pre-aligned position. For this case, the user can specify AsIs as a starting position.

And finally, there is a method to generate N random starting positions where N is user-defined as well as the maximum translation allowed between random starts. In most cases, the default Inertial frame starting positions are completely sufficient to find optimal overlap.

Note that the fit object is not moved during the optimization. Part of the results returned from the calculation are the transforms necessary to move the fit object into the final orientation. This allows the user to skip this step if only scores are desired. It also allows application of the same transform to other objects.

				Shape Toolkit Java Programming Manual

				Shape Toolkit Java Programming Manual