I would like to alter the activity of an enzyme towards a novel substrate.
I have a recently solved methyltransferase structure, which uses S-adenosylmethionine as a co-factor. There is also a substrate bound in the correct position for reaction. The part of that molecule which receives the methyl group is the same as in my novel substrate, but there are some major differences in the rest of the substrate. I can easily model the novel substrate into the active site in the conformation required to give the desired product, so I don't think that it is necessary to generate a transition state model.
What would be your recommendations to tackle this problem? I am thinking that the catalytic part of the active site (the SAM, the novel substrate, and the constellation of residues around the SAM and the conserved part of the substrate should be fixed, and only the positions around the altered part of the substrate should be re-designed. Does this seem reasonable?
Sounds reasonable to me.
One suggestion I have is to expect to take an iterative approach. Set up a design run, generate a low-mid number of output designs (probably 100-200), filter out the obviously bad ones, and then examine those structures. Look for what's working, what's not. How might the designs be improved? With the answers to those questions you can start formulating how you would change the design protocol to make better structures. This may involve changing which residues are allowed to repack or redesign, or it may involve adding in constraints during the design to keep various components where they should be.
This is normally the way these sorts of design runs work. You do a rather simple design run, discover there's an obvious problem, fix the problem and then repeat. If you're lucky, after about a half dozen or so rounds of correcting mistakes and re-running you should get results which look decent.