RNAfcg: RNA Flexibility Prediction Based on Topological Centrality and Global Features.

Area of Science:

• Computational Biology
• Bioinformatics
• Molecular Biology

Background:

• RNA dynamics are intrinsically linked to their biological functions.
• Predicting RNA flexibility is crucial for understanding RNA behavior, but experimental methods are laborious.
• Developing accurate theoretical methods for RNA flexibility prediction is essential.

Purpose of the Study:

• To develop an effective machine learning method, RNAfcg, for predicting RNA flexibility.
• To identify key features that determine RNA flexibility and dynamics.
• To provide a faster and more reliable alternative to experimental methods for RNA flexibility assessment.

Main Methods:

• Utilized a Random Forest (RF) machine learning model.
• Incorporated novel features: topological centralities, flexibility-rigidity index, and global characteristics.
• Combined these with traditional sequence and structural features for training.

Main Results:

• The novel topological, flexibility-rigidity, and global features significantly contribute to RNA flexibility prediction.
• Topological features demonstrated the most substantial impact, highlighting the importance of structural topology.
• RNAfcg achieved a Pearson correlation coefficient (PCC) of 0.6619, outperforming state-of-the-art machine learning and Gaussian Network Model (GNM) methods.

Conclusions:

• RNAfcg is a highly effective tool for predicting RNA flexibility and understanding RNA dynamics.
• The study underscores the critical role of structural topology in determining RNA flexibility.
• This method aids in predicting RNA function and offers a valuable computational approach for RNA research.