Summary
Accurate fitness landscapes are unnecessary for productively engineering enzymes or proteins. This has been true empirically when generating sequences using random techniques such as error-prone PCR, but (1) demonstrated that autonomous laboratories that design sequences using Gaussian processes can iteratively improve fitness even in stages where the algorithm is not modeling landscapes accurately.
See also
1.
Rapp JT, Bremer BJ, Romero PA. Self-driving laboratories to autonomously navigate the protein fitness landscape. Nature Chemical Engineering. 2024;1(1):97–107. Available from: https://doi.org/10.1038/s44286-023-00002-4