EDIT: Apparently this is called reactive molecular dynamics. It seems that the ReaxFF potential function is used for some reactions like this.
I am interested in adding support for reactions to a coarse grained molecular dynamics simulations. On the whole each particle will represent a large biomolecule.
Most of the literature I have surveyed indicates that the force fields used in molecular dynamics make it unsuitable for modelling reactions (as in determining if a reaction will take place). For my simulation I don't think that I need to determine if it is electronically/energetically possible for it to occur. I want to simulate rule-based reactions.
I have seen one book where the author says that with prior knowledge it is possible to assume that if two particles in your simulation collide they will react.
1) Is this correct? If so, what would one need to know about the particles? I know that Na$^+$ and Cl$^-$ react to form NaCl. Could I just detect if they collided and then replace them with an NaCl simulation particle?
2) Would this work with large molecules (one simulation particle per molecule)? If I knew that enzyme A acted on substrate B to produce substrates C and D (A+B$\rightarrow$A+C+B), then if I detected a collision in my molecular dynamics simulation would it be scientifically correct to perform the reaction?
3) What prior knowledge would I need for this to work? Is this the sort of knowledge that is readily available? (I.e. A+B$\rightarrow$C 50% of the time)
4) Is this good science? Is it unlikely that a simulation of a biological system using this rule based technique would get published in a good journal?