RNA is often thought of as a simple messenger that carries genetic information, but many RNA molecules actually fold into complex three dimensional shapes that allow them to perform important biological functions. These structures influence how RNA interacts with proteins, regulates genes, and participates in disease processes. Understanding RNA structure is therefore critical for fields like drug development, diagnostics, and synthetic biology.

A new tool called trRosettaRNA aims to make this process much easier. Developed by researchers at Shandong University, the platform uses deep learning to predict the three dimensional structure of RNA molecules directly from their nucleotide sequence. Instead of relying only on slow experimental methods, scientists can now generate structural models in about an hour for typical RNA sequences.

Overview of trRosettaRNA protocol

Starting with the sequence of a target RNA, the protocol generates a MSA and a SS. Optionally, users can also input their custom MSA and/or SS. The generated/input MSA and SS are fed into an end-to-end neural network to predict initial 3D atomic coordinates and 2D internucleotide distances. Subsequently, the predicted coordinates are then optimized through fast relaxation or energy minimization (min.) (depending on the severity of steric clash) to produce a physically realistic structure. The final outputs include the predicted 3D structure and an associated confidence score (pLDDT). Clashscores can be independently checked using programs such as Molprobity

The system works by taking an RNA sequence as input and running it through a neural network trained to recognize patterns in how RNA folds. Users can also include additional information, such as sequence alignments or known secondary structures, to improve prediction accuracy. After the initial prediction, the model applies an energy optimization step to refine the structure and correct any unrealistic features.

One of the key advantages of trRosettaRNA is that it is available as a web-based server, making it accessible to researchers without advanced computational expertise. It also offers strong visualization tools, allowing users to explore predicted RNA structures in detail. For more advanced users, a standalone version is available, which allows integration into automated workflows and ensures data privacy for sensitive sequences.

The tool has already demonstrated strong performance in predicting structures for RNA families that previously lacked experimental data. In several cases, the predicted models closely matched structures that were later confirmed experimentally, highlighting the reliability of this approach.

As RNA becomes an increasingly important target in therapeutics, including mRNA vaccines and RNA-based drugs, tools like trRosettaRNA are helping bridge the gap between sequence data and structural insight. By combining RNA sequencing data with advanced modeling, researchers can better understand RNA function and accelerate the development of new therapies.

Availability – The web server and standalone package of trRosettaRNA are available at https://yanglab.qd.sdu.edu.cn/trRosettaRNA/ and https://github.com/YangLab-SDU/trRosettaRNA2, respectively.