Machine Learning Ensemble Directed Engineering of Genetically Encoded Fluorescent Calcium Indicators.

Area of Science:

• Biotechnology
• Protein Engineering
• Computational Biology

Background:

• Genetically encoded fluorescent indicators (GEFIs) are vital for monitoring biological activity but are challenging to optimize using traditional methods.
• Trial-and-error mutagenesis is time-consuming and inefficient for complex protein sensors like GEFIs.
• Developing improved GEFIs is crucial for advancing real-time biological research and diagnostics.

Approach:

• Applied machine learning to predict functional outcomes of protein sensor mutations.
• Developed an ensemble of three regression models trained on experimental GCaMP mutation data.
• Performed in silico screening of 1423 novel GCaMP variants to identify promising candidates.

Key Points:

• Identified novel ensemble-derived GCaMP (eGCaMP) variants with faster kinetics and larger fluorescent responses.
• Discovered eGCaMP2+, a combinatorial mutation outperforming existing GCaMP generations in dynamic range.
• Demonstrated machine learning's efficacy in pre-screening mutants and accelerating discovery.

Conclusions:

• Machine learning significantly enhances the efficiency of protein engineering for GEFIs.
• This study validates machine learning as a powerful tool to overcome limitations in traditional mutagenesis.
• The developed eGCaMP variants offer improved capabilities for biological sensing applications.