Machine Learning Informs RNA-Binding Chemical Space.

Area of Science:

• Medicinal Chemistry
• Chemical Biology
• Computational Chemistry

Background:

• Targeting RNA with small molecules is a promising area in drug discovery.
• Understanding the chemical space of RNA binders is currently limited compared to protein binders.

Purpose of the Study:

• To create a comprehensive, publicly available library of RNA-binding small molecules.
• To apply machine learning for characterizing RNA-binding molecules and informing future drug design.

Main Methods:

• Utilized small molecule microarray (SMM) screening to test 24,572 small molecules against nucleic acids.
• Performed 36 individual nucleic acid SMM screens, assaying over 1.6 million interactions.
• Employed machine learning algorithms to build predictive models for RNA-binding molecules.

Main Results:

• Identified a set of 2,003 RNA-binding small molecules.
• Established the Repository Of BInders to Nucleic acids (ROBIN) library, the largest public dataset of its kind.
• Developed highly predictive and interpretable machine learning models for RNA binders.

Conclusions:

• The ROBIN library significantly expands the known chemical space for RNA-targeted therapeutics.
• Machine learning applied to experimental screening data is effective for characterizing RNA-binding molecules.
• This approach accelerates the discovery and development of novel RNA-targeted drugs.