Fitness functions for RNA structure design

Area of Science:

• Computational biology
• Molecular biology
• Bioinformatics

Background:

• RNA design algorithms are essential for engineering RNA therapeutics.
• Fitness functions guide these algorithms, but their comparative performance is understudied.
• Previous comparisons of RNA design fitness functions were published nearly 20 years ago.

Purpose of the Study:

• To survey current RNA design approaches, focusing on fitness functions.
• To experimentally compare the most widely used RNA design fitness functions.
• To identify the most effective fitness function for RNA sequence design.

Main Methods:

• Surveyed existing RNA design algorithms and their fitness functions.
• Conducted experimental comparisons of fitness functions using synthetic and natural RNA sequences.
• Evaluated fitness functions based on their ability to predict or generate target RNA structures.

Main Results:

• Maximizing the probability of a target structure significantly outperforms minimizing ensemble defect.
• The probability metric aligns better with naturally evolved RNA sequences and structures.
• Minimizing structure distance to the minimum free energy prediction was found to be an ineffective fitness function.

Conclusions:

• Maximizing probability is the most effective fitness function for RNA sequence design.
• This finding has significant implications for the development of RNA-based therapeutics.
• Current trends in RNA design, such as minimizing structure distance, may be suboptimal.