Reliable method for predicting the binding affinity of RNA-small molecule interactions using machine learning.

Area of Science:

• Computational Biology
• Drug Discovery
• Molecular Biology

Background:

• Ribonucleic acids (RNAs) are crucial for cellular regulation, and their dysregulation is linked to various human diseases.
• RNAs are increasingly recognized as potential drug targets for therapeutic interventions.
• Identifying novel RNA targets and their small molecular inhibitors is essential for advancing disease treatment.

Purpose of the Study:

• To develop machine learning models for predicting the binding affinity of small molecules to six specific RNA subtypes.
• To accelerate the identification of disease-associated RNA targets and small molecule inhibitors.
• To provide a freely accessible web server for these predictive models.

Main Methods:

• Development of machine learning models tailored for six RNA subtypes: aptamers, miRNAs, repeats, ribosomal RNAs, riboswitches, and viral RNAs.
• Analysis of RNA sequence composition, flexibility, and the polar nature of RNA-binding ligands as key predictive features.
• Evaluation using jack-knife tests and validation with external blind test datasets.

Main Results:

• The developed models achieved an average Pearson correlation (r) of 0.83 and a mean absolute error of 0.66.
• The models demonstrated reliability even with limited data for several RNA subtypes.
• Performance surpassed existing quantitative structure-activity relationship (QSAR) models on external validation datasets.

Conclusions:

• Machine learning models can reliably predict RNA-small molecule binding affinity.
• These models are valuable tools for accelerating the discovery of RNA-targeted therapeutics.
• The RNA-Small molecule binding Affinity Predictor web server is available for public use.