I have an X-ray structure of an enzyme with reported activity to a small molecule. This activity is rather low since it is not the native substrate.

I can run molecular modeling simulations (e.g. using DOCK) to estimate the enzyme's binding behavior to different compounds. However, what I would like to be able to do is to mutate the binding pocket so that it better binds my small molecule of interest.

Exploring all possible amino acid substitutions is computationally too expensive, so I am wondering whether smarter ways of looking at this problem have been developed.

Can someone point me in the right direction? I've searched Google but can't really find a straight forward answer

  • $\begingroup$ Could you choose rational mutations that you think may facilitate binding and only study those? The results of one mutation may inform the next. $\endgroup$
    – canadianer
    Feb 10, 2017 at 2:10
  • $\begingroup$ @canadianer There is probably a way. However, I would not want to have to reinvent the wheel if someone has done something like that before $\endgroup$
    – Dahlai
    Feb 14, 2017 at 3:01

1 Answer 1


You can try an evolutionary algorithm. If you can quickly evaluate the binding affinity you can initialize a set of copies of your protein but with random mutations. You can assign them a fitness score based on the affinity of the binding. The ones with highest affinity you replicate again with some probability of mutating each amino acid and you continue for as many generations as you need. If you find your parameters (mutation rate, population size, etc) correctly you can easily optimize the binding without knowing anything a priori about the different sites. There is a lot of literature on this, I suggest you do a quick Google scholar search on directed evolution or evolutionary algorithms to optimize binding affinity.

Paper on in silico evolution to optimize protein-protein binding

  • $\begingroup$ Thanks for the input. If I understand correctly they used in silico design and experimental optimization to get to the final result. They did not use an evolutionary algorithm. However, I think that this is generally a promising path to explore. I am just wondering whether so already developed such a tool. $\endgroup$
    – Dahlai
    Feb 14, 2017 at 3:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.