I'm debating moving from (M06-2X:PM7:UFF) to (M06-2X:PM7:AMBER) for modelling a reaction mechanism in an enzyme. I've already examined the errors introduced by my partitioning and whether they cancel sufficiently.
However, as dispersion/H-bonding is quite vital for describing this mechanism, I'm tempted to switch from UFF to AMBER. However, due to the way ONIOM works (it does an MM calc for the entire system, a SE calc for the mid + high layers and a DFT for the high layer), I'll need to parameterize my ligands due to them not being available in Gaussian16's Amber parameter bank. For the cofactor, there already exist online parameters for the product and reactant structures alike and the hypothesized mechanisms do not have intermediates for the cofactor, only the substrate.
For the substrate, no cofactors exist. In literature, people parameterized this substrate using RESP charges obtained through DFT calculations - something I could do myself. There would be the reactant and product structures that need parameterization, and I am not sure if I also need to parameterize the intermediate carbocation (reaction path 1) or the intermediate oxyanion (reaction path 2).
My supervisor is not too familiar with AMBER, and I don't want to bother him with questions that probably lead nowhere. His past publications with ONIOM used UFF as the low layer. Myself I'm mostly familiar with the theory of HF, MPn, B3LYP, Minnesota functionals and coupled cluster methods.
My hunch is that I can't use AMBER as low-layer in ONIOM to model transition states or intermediates. However, I'd rather ask those who may have done so before giving up on this thought.
