Machine Learning-Assisted Engineering of Light, Oxygen, Voltage Photoreceptor Adduct Lifetime.

Area of Science:

• Biochemistry
• Molecular Biology
• Synthetic Biology

Background:

• Flavin-binding, blue-light-sensing, light-oxygen-voltage (LOV) photoreceptors are crucial for biological process control.
• Engineering LOV photoreceptors with tailored kinetics is vital for applications like optogenetics and plant biomass production.
• The dark-recovery step, involving adduct scission, is rate-limiting and challenging to engineer due to its long timescale and complexity.

Purpose of the Study:

• To address challenges in engineering LOV photoreceptor dark recovery kinetics.
• To develop a machine learning approach for predicting LOV variants with fine-tuned dark recovery.
• To expand the utility of LOV domains in synthetic (opto)biology.

Main Methods:

• Utilized machine learning (ML) trained on limited literature data.
• Employed an iterative approach of data generation and experimental validation.
• Designed and tested LOV domain variants with altered dark recovery properties.

Main Results:

• Successfully predicted LOV domain variants with significantly modified dark recovery.
• Achieved a 7-orders-of-magnitude span in adduct-state lifetimes for engineered LOV variants.
• Demonstrated ML's capability to guide protein design with scarce data and no mechanistic model.

Conclusions:

• Machine learning is a viable strategy for protein engineering, particularly for complex processes like LOV photoreceptor kinetics.
• This work provides optimized LOV tools for synthetic (opto)biology applications.
• The ML approach overcomes limitations of traditional methods for engineering protein function.