Engineering highly active nuclease enzymes with machine learning and high-throughput screening.

Area of Science:

• Synthetic biology
• Protein engineering
• Bioinformatics

Background:

• Enzyme optimization for new environments is crucial in synthetic biology.
• Challenges include rugged fitness landscapes and expensive experiments.
• Nuclease enzymes can degrade biofilms in chronic wounds.

Purpose of the Study:

• To develop a machine learning (ML) framework, TeleProt, for designing diverse protein libraries.
• To improve the catalytic activity of a nuclease enzyme for biofilm degradation.
• To overcome limitations of traditional directed evolution methods.

Main Methods:

• TeleProt framework blending evolutionary and experimental data.
• High-throughput experimental screening of enzyme variants.
• Application to a nuclease enzyme targeting chronic wound biofilms.
• Comparison with directed evolution (DE).

Main Results:

• TeleProt identified superior top-performing enzymes compared to DE.
• Achieved a higher hit rate for diverse, high-activity variants.
• Successfully designed a high-performance initial library without prior data.
• Released an extensive dataset of 55,000 nuclease variants.

Conclusions:

• TeleProt offers a powerful ML-guided approach for protein design.
• It surpasses directed evolution in enzyme optimization efficiency and diversity.
• The released dataset will advance ML-driven enzyme engineering.