For decades, researchers have struggled to map the complex 3D structures of RNA—a biomolecule critical to life and a promising drug target. The process of experimentally determining RNA structures is slow and labor-intensive, leaving a significant gap in the available data. At the current pace, it could take years to unlock the structural secrets of disease-related RNA molecules. However, a new computational breakthrough may change that timeline dramatically.

A research team at Purdue University, led by Daisuke Kihara, has developed NuFold, a machine-learning-powered tool designed to model RNA structures with unprecedented speed and accuracy. Their findings, published in Nature Communications, suggest that NuFold could revolutionize RNA research in the same way AlphaFold did for proteins—potentially expediting medical discoveries by decades.

NuFold: A Game-Changer for RNA Structural Biology

Unlike proteins, which have been extensively studied for their 3D structures, RNA remains largely uncharted territory. Yet, understanding RNA’s intricate folding is crucial, particularly as scientists explore its role in gene regulation, disease mechanisms, and therapeutic development.

“NuFold is the RNA equivalent of AlphaFold,” says Kihara. “AlphaFold was a breakthrough in computational protein structure prediction and won the Nobel Prize in Chemistry in 2024. Our goal is to extend that innovation to RNA, which has historically been much more challenging to model.”

A key advantage of NuFold is its ability to predict RNA structures from sequence data with remarkable accuracy. Traditional methods rely on energy-based modeling, which can be computationally expensive and limited in precision. NuFold, on the other hand, uses machine learning techniques to capture RNA’s inherent flexibility, outperforming conventional approaches in benchmark tests.

The tool is open-source and publicly available, allowing researchers worldwide to leverage its capabilities. With a Google Colab notebook provided for ease of use, even scientists without deep computational expertise can access NuFold for their studies.

Bridging the Gap in RNA Drug Discovery

The applications of NuFold extend beyond academic curiosity. RNA-based drugs—such as mRNA vaccines and gene therapies—are gaining traction, but their development is hindered by the lack of structural data. By making accurate RNA structure predictions accessible, NuFold could accelerate the discovery of new treatments for diseases linked to RNA, including neurodegenerative disorders and viral infections.

“It took over three years to develop NuFold,” says Yuki Kagaya, a postdoctoral research assistant and the tool’s main developer. “One of its key features is how it represents RNA internally—accounting for base pairs that are crucial to structure while maintaining flexibility. In benchmark tests, it has outperformed energy-based methods and even recent deep-learning approaches in local structure prediction.”

By enabling researchers to visualize RNA structures, NuFold can help scientists design new therapeutics, identify drug-binding sites, and predict how mutations affect RNA function. This breakthrough has implications not only for drug development but also for understanding fundamental biological processes.

A Team Effort Driving Innovation

The development of NuFold was a collaborative effort within Purdue’s Structural Biology Group, bringing together experts in biological sciences and computer science. Kihara, the lead author of the study, directed the project alongside Kagaya, who spearheaded the coding. Other contributors include Zicong Zhang, Nabil Ibtehaz, Xiao Wang, Tsukasa Nakamura, and Pranav Deep Punuru, who worked on coding, benchmarking, and developing the web server interface.

The research was supported by Purdue University’s Rosen Center for Advanced Computing (RCAC), the NSF XSEDE (now ACCESS) program, and the Oracle for Research Cloud Grant.

“To solve problems that cannot be immediately addressed through experiments, we developed NuFold as a computational solution,” says Kihara. “By predicting RNA’s 3D structures from sequence data, we hope to bridge the gap in RNA research and provide a powerful tool for the scientific community.”

As NuFold continues to evolve, it has the potential to reshape the landscape of RNA research—turning years of experimental work into mere computational predictions and unlocking new possibilities in medicine and biotechnology.