DeCOIL: Optimization of Degenerate Codon Libraries for Machine Learning-Assisted Protein Engineering.

Area of Science:

• Biotechnology
• Computational Biology
• Protein Engineering

Background:

• Machine learning (ML) advances protein engineering by proposing protein variants.
• Synthesizing large variant libraries is a costly bottleneck.
• Degenerate codon (DC) libraries offer a cost-effective mutagenesis strategy.

Purpose of the Study:

• To develop an optimized computational method for designing DC libraries tailored for ML-assisted protein engineering.
• To enable the efficient exploration of protein variants with high fitness and sequence diversity.

Main Methods:

• Introduced DEgenerate Codon Optimization for Informed Libraries (DeCOIL), a generalized method for DC library optimization.
• Validated DeCOIL using computational simulations and experimental wet-lab studies.
• Developed accompanying open-source software for DeCOIL implementation.

Main Results:

• DeCOIL effectively optimizes DC libraries for ML-assisted protein engineering.
• The method demonstrated success in two distinct case studies.
• DeCOIL enables sampling of protein variants with desired high fitness and diversity.

Conclusions:

• DeCOIL addresses limitations of existing methods for DC library design in protein engineering.
• The method is direct, user-friendly, generalizable, and scalable.
• DeCOIL facilitates the generation of informed libraries for exploring protein sequence space efficiently.