Our group has measured Kcat on a WT enzyme and 28 single-AA substitution variants. How would one go about using this information to predict the effect of other mutations on its activity?

For context, I'm a statistician and have just recently joined the group. I've seen a bunch of server blades in the backroom so I'm assuming we have the resources to attempt most computational+statistical methods.

From my initial review of the literature, sequence-based methods require a much larger sample size and structure-based methods would give us physicochemical properties of the enzyme-ligand complex. From that, we could predict activity, but I don't see how our in vitro data could be added to refine the prediction.