An AI-guided framework reveals conserved features governing microRNA strand selection

Area of Science:

• Molecular Biology
• Genetics
• Bioinformatics

Background:

• MicroRNAs (miRNAs) are key gene expression regulators.
• The mechanism of miRNA strand selection (5p vs. 3p) during biogenesis is not fully understood.

Purpose of the Study:

• To develop a framework for understanding miRNA strand selection logic.
• To build a predictive model for miRNA strand preference across species.

Main Methods:

• Developed a high-throughput platform for quantifying miRNA strand usage in *Caenorhabditis elegans*.
• Created an AI-driven machine learning model integrating 77 features to predict strand preference.
• Validated the model across nematodes and vertebrates, including humans.

Main Results:

• Identified conserved, context-dependent rules governing miRNA strand selection.
• Revealed compositional and structural biases in strand preference that are conserved and functionally repurposed.
• Demonstrated that strand selection is not stochastic but follows predictable patterns.

Conclusions:

• Established the first unified, generalizable model for miRNA strand selection.
• Combined large-scale experimentation with AI to uncover a programmable layer of gene regulation.
• Provided open-access resources for the research community.