Toehold-VISTA: A machine learning approach to decipher programmable RNA sensor-target interactions.

Area of Science:

• Synthetic biology
• Molecular diagnostics
• Computational biology

Background:

• RNA-based biosensors are crucial for synthetic biology and diagnostics, but their design is time-consuming.
• Understanding RNA-RNA interactions and structure-function relationships is key to improving sensor performance.
• Current methods for RNA sensor design are slow and lack predictive power.

Purpose of the Study:

• To present a machine learning-guided framework, VISTA, for rapid and versatile in-silico RNA-targeting analysis.
• To accelerate the design and engineering of high-performance RNA biosensors.
• To improve RNA sensor function through a target-aware design strategy.

Main Methods:

• VISTA integrates biophysical modeling of sensor and target RNAs with partial least squares discriminant analysis (PLS-DA).
• High-throughput experimental measurements and sequence-structure feature extraction were used to train predictive models.
• Toehold switches were employed as a model RNA sensor system to validate the VISTA framework.

Main Results:

• VISTA successfully captured key determinants of RNA sensor performance.
• Toehold-VISTA demonstrated the ability to design RNA sensors with improved function against SARS-CoV-2 RNA.
• The framework enables rapid, target-aware design of RNA sensors.

Conclusions:

• VISTA provides a broadly applicable strategy for accelerating RNA sensor engineering.
• This machine learning approach enhances the development of RNA-based tools for biotechnology and diagnostics.
• The study establishes a foundation for more efficient design of RNA biosensors.