docking predict free binding energy that can be used in several ways to predict the affinity (Ki) , but how to detect whether this binding will cause activation or inhibition of the enzyme?
If a set of known activators and inhibitors and activators exist, the best strategy may be to ignore the protein and look for chemical similarities and differences within the inhibitor and activator sets using an abstract representation of each molecule. This type of procedure is known to be very good (generally more accurate than docking) for screening databases for biologically active compounds once one active is known.
This can be done in the free program OpenBabel by the commands
babel query.sdf known_actives.sdf -ofpt
babel query.sdf known_inhibitors.sdf -ofpt
this will give the chemical similarity between the molecule you want (query.sdf) and each molecule in the set of either known actives or inhibitors. A high number of matches between one set and not the other will likely tell you which class it falls into.
This can be made more quantitative by constructing what is known as a ligand profile which is a representation of how often each chemical feature appears in the set. However, I am not not sure if this can be done directly with standard off the shelf programs.
The enzyme inhibitor option on patchdock will not distinguish allosteric activators from inhibitors. It is just used to change the parameters for small molecule and antibody-antigen vs other types