Advances in machine learning for directed evolution.

Area of Science:

• Biotechnology
• Computational Biology
• Protein Engineering

Background:

• Directed evolution is crucial for protein engineering but relies on expensive experimental screening.
• Machine learning (ML) offers in silico screening to accelerate this process.
• Acquiring sufficient sequence-function data for ML model training remains a significant bottleneck.

Purpose of the Study:

• To review recent advances in ML for directed evolution.
• To highlight methods that leverage protein sequence data to reduce or eliminate the need for extensive sequence-function data.
• To discuss ML strategies for generating novel functional protein sequences and exploring protein sequence space.

Main Methods:

• Review of ML approaches applied to protein sequence data.
• Analysis of methods augmenting limited sequence-function data with raw sequence data.
• Examination of generative ML models for protein sequence diversity.

Main Results:

• ML models trained on protein sequences can effectively augment limited experimental data.
• Strategies exist to minimize or remove the requirement for costly sequence-function data in ML-driven directed evolution.
• Generative ML models enable efficient exploration of protein sequence space for novel functions.

Conclusions:

• ML, particularly using readily available protein sequence data, significantly enhances directed evolution efficiency.
• Future efforts should focus on developing and refining ML models that minimize experimental data requirements.
• Generative models hold promise for discovering new protein functionalities by exploring vast sequence landscapes.