Probing the Specificity of Fluorescent Deoxyribozymes Using Single-Step Selections and Machine Learning.

Area of Science:

• Biochemistry
• Molecular Biology
• Synthetic Biology

Background:

• Proteins and nucleic acids form specific binding sites crucial for biological functions.
• Modulating biochemical specificity is vital for enzyme engineering and drug design.

Purpose of the Study:

• To systematically investigate the specificities of self-phosphorylating deoxyribozymes.
• To identify mutations that alter deoxyribozyme specificity using selection and machine learning.

Main Methods:

• Biochemical assays were used to test deoxyribozyme activity with various substrates.
• Single-step selections were performed on a deoxyribozyme library to isolate variants with altered specificity.
• Machine learning models were developed to predict the effects of mutations on specificity.

Main Results:

• Deoxyribozymes initially showed high specificity for the coumarin substrate 4-MUP.
• Selected variants reacted with both 4-MUP and the similar substrate diFMUP.
• Four specific mutations were identified that modulate the deoxyribozyme's specificity, confirmed by assays and machine learning predictions.

Conclusions:

• Single-step selections are effective for identifying mutations that change deoxyribozyme specificity.
• Machine learning can successfully model complex data from in vitro selection experiments.
• This work provides insights into engineering enzyme specificity for biotechnological applications.