This creates a new class to be used for assessing or scoring models with SOAP-Loop.
For assessing models (i.e., a one-time score for each model after optimization is complete), simply assign an instance of this class to AutoModel.assess_methods or LoopModel.loop.assess_methods (see Section 2.3.3 for an example). For manual assessment, see Selection.assess().
To add to the energy function (e.g., to use in optimization, which is not usually what you want), you must add an instance of this class to the relevant EnergyData.energy_terms list, in the same way as for user-defined energy terms (see Section 7.1.3). (Note that the default value of 4.0 for EnergyData.contact_shell is too small to be used with SOAP scores. Set it to the undefined value (-999) to be sure not to discard SOAP statistics for longer distances.)
library is the name of the SOAP-Loop library. This library is not included with the MODELLER distribution due to size (it is roughly 500MB) but can be downloaded from the SOAP web site.
SOAP-Loop is an orientation-dependent potential. It can only reliably be used for scoring (not optimization) as its first derivatives are zero.