An AI-Guided Framework Reveals Conserved Features Governing microRNA Strand Selection

Area of Science:

• Molecular Biology
• Genetics
• Bioinformatics

Background:

• MicroRNAs (miRNAs) are key regulators of gene expression.
• The mechanism determining which miRNA strand (5p or 3p) is preferentially used during biogenesis is not fully understood.
• Understanding miRNA strand selection is crucial for deciphering gene regulation.

Purpose of the Study:

• To develop a comprehensive framework for understanding miRNA strand selection.
• To identify the molecular rules governing the choice between miRNA 5p and 3p strands.
• To create a predictive model for miRNA strand usage across different species.

Main Methods:

• Developed a high-throughput quantitative PCR (qPCR) platform for precise miRNA strand quantification in *Caenorhabditis elegans*.
• Constructed a machine learning model trained on experimentally validated strand usage data.
• Integrated 77 biologically informed features into the AI model for prediction.

Main Results:

• The developed framework accurately predicts miRNA strand preference in *C. elegans* and vertebrates, including humans.
• Identified conserved compositional and structural biases influencing strand selection across species.
• Demonstrated that miRNA strand selection is a regulated, context-dependent process, not random.

Conclusions:

• This study presents the first unified, generalizable model for miRNA strand selection.
• The findings reveal a conserved, programmable layer of gene regulation governed by specific rules.
• Open-access resources, including a database and predictive tools, are provided to the research community.