Pusan National University team develops AI tool to map RNA binding sites

Researchers led by Professor Giltae Song at Pusan National University developed circMAC, an AI tool that predicts candidate microRNA (miRNA) binding locations across full-length circular RNA (circRNA) molecules.

Unlike approaches that may treat RNA sequences as linear, circMAC preserves the circular sequence’s continuity, including at the back-splice junction, and accounts for the specific miRNA being studied. It assigns a binding probability to each circRNA nucleotide.

In the researchers’ benchmarks, circMAC outperformed the evaluated RNA language models and rule-based methods. At one reported test setting, it achieved an F1-macro score of 0.762 and an AUROC of 0.902.

The team says the tool is intended to help researchers prioritize candidate sites for laboratory testing—not to prove that an interaction is functional or establish a clinical application.