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

Tempe, AZ — January 14, 2026 MicroRNAs, whose discovery was recognized with the 2025 Nobel Prize in Physiology or Medicine, are central regulators of gene expression, yet a fundamental question has remained unanswered: how cells choose between the two strands produced from each microRNA precursor. A new study published in Nucleic Acids Research reveals that this decision follows conserved, programmable rules rather than chance.

Led by Marco Mangone, professor in the Center for Personalized Diagnostics at the Biodesign Institute and the School of Life Sciences at Arizona State University, the research team combined large-scale experiments with artificial intelligence to decode the logic of microRNA strand selection. Using the model organism Caenorhabditis elegans, the team developed a high-throughput method called HiTmiSS to precisely measure strand usage across development and in specific tissues.

The researchers generated thousands of strand-specific measurements and in collaboration with Dr. Heewook Lee, assistant professor in the School of Computing and Augmented Intelligence at Arizona State University, used these data to train a machine-learning model that integrates 77 biologically informed features of microRNA sequence and structure. The model accurately predicts which strand is selected, not only in nematodes but also across vertebrates, including humans.

“Our results show that strand selection is not random,” said Mangone. “It follows conserved rules encoded in the microRNA duplex itself, but those rules can be tuned by developmental and tissue context.”

The study also reveals how evolution has reshaped strand selection. While many invertebrate microRNAs retain flexible strand usage, mammalian microRNAs show a strong bias toward a single dominant strand, suggesting increased regulatory precision in more complex organisms.

“This work provides the first unified, experimentally grounded framework for understanding microRNA strand selection,” said Dalton Meadows, first author of the study. “By combining biology with AI, we can now predict strand usage and understand how it evolves.”

To support the research community, the authors have released all experimental data, prediction scores, and analysis tools as open-access resources, enabling strand-selection prediction across species.

Together, the findings establish microRNA strand selection as a regulated layer of gene control and demonstrate how artificial intelligence can uncover hidden biological rules conserved across evolution.