The effective design of cold-start enzyme libraries to balance fitness and diversity enables access to enzyme variants that are readily evolvable and close to the optima in the fitness landscape. Here, the authors develop MODIFY (machine learning-optimized library design with improved fitness and diversity), a machine learning algorithm to co-optimize expected fitness and sequence diversity of starting libraries, enhancing the efficiency of directed evolution in enzyme engineering.
- Kerr Ding
- Michael Chin
- Yunan Luo
