RNA molecules aren’t just messengers carrying genetic information, they’re also finely tuned with chemical modifications that affect how cells function. Understanding these modifications, often called the “epitranscriptome,” can reveal important insights into gene regulation, disease processes, and potential new therapies.

Researchers from the University of Mainz in Germany have developed a powerful new tool called ModiDeC, short for Modification Detection Classifier. This deep-learning model can identify and distinguish between several common RNA modifications, including N6-methyladenosine, inosine, pseudouridine, 2′-O-methylguanosine, and N1-methyladenosine. What makes ModiDeC especially valuable is that it works with direct RNA sequencing using nanopore technology, which reads RNA molecules directly rather than converting them to DNA first.

To train and validate ModiDeC, the team created a large reference database using both synthetic RNA and samples from human cells and blood. The tool performed with high accuracy across different types of samples and was able to differentiate between modifications even when they occurred in similar RNA sequence patterns.

ModiDeC is designed for accessibility. It includes a graphical user interface and an Epi2ME pipeline, making it easier for researchers to apply it to their own datasets and even customize it for detecting new types of RNA modifications. Because of its accuracy and reproducibility, ModiDeC is poised to become a key player in decoding the complex language of RNA.

Availability – The neural network and the GUI codes for ModiDeC are available at Zenodo (DOI:10.5281/zenodo.15719181) and GitHub: https://github.com/mem3nto0/ModiDeC-RNA-modification-classifier