Tuning the Performance of Synthetic Riboswitches using Machine Learning

Area of Science:

• Synthetic biology
• Molecular biology
• Bioinformatics

Background:

• Riboswitches are crucial genetic regulatory elements used in clinical, technological, and synthetic biology.
• Optimizing riboswitch regulatory behavior is essential for enhancing their functionality and applications.

Purpose of the Study:

• To improve the regulatory behavior of a tetracycline (tc)-dependent riboswitch using a machine learning approach.
• To develop a bioinformatics model for predicting and optimizing the performance of tandem riboswitches.

Main Methods:

• A machine learning model combining random forest analysis and a convolutional neural network was developed.
• The model predicted the switching behavior of tandem tc-aptamer riboswitches.
• Biophysical parameters and hydrogen bond patterns were analyzed for their influence on regulation.

Main Results:

• The machine learning approach significantly improved the tc riboswitch device's regulation.
• The dynamic range of the tc riboswitch was extended from 8.5-fold to 40-fold.
• Both biophysical parameters and hydrogen bond patterns were identified as key factors influencing riboswitch regulation.

Conclusions:

• The novel machine learning method successfully optimized the tc-dependent riboswitch.
• This approach offers a promising strategy for the development and optimization of other riboswitches.
• The enhanced riboswitch device has broad potential in synthetic biology and clinical applications.