Researchers from The University of Texas at Austin and Sanofi have developed a breakthrough model. It is for simulating the task of designing new mRNA strands which can be used in advanced COVID-19 vaccines. It is only the second such method in the world. The breakthrough, described in two recent papers in Nature Biotechnology, could pave the way for a new kind of medicine to treat diseases: messenger RNA. Scientists using RiboNN aim to reduce trial and error and speed up development by getting closer to nailing exactly which tissues or organs are the best to target.
What RiboNN Brings to the Prediction of Protein Production
This roadmap for proteins is encrypted in mRNA, and it’s often the bottleneck in developing therapies to get cells to make enough of the right protein. RiboNN was trained with >10,000 experiments coming from more than 140 human and mouse cell types to have a wide basis in learning patterns of translation efficiency. RiboNN, developed jointly by experts in UT Austin and Sanofi’s artificial intelligence programs, is designed to outperform other existing models at predicting which mRNA sequences will be most effective in certain types of cells.
In tests, RiboNN proved to be almost two times more accurate in the prediction of translation efficiency than existing methods. What makes RiboNN particularly appealing, however, is its promise for informing sequence design from an early stage. By predicting which mRNA variants are more likely to result in greater protein output. Or be more effective for a target organ, such as the liver or the lung.
Implications for Cancer, Genetic Disorders, and Precision Medicine
RiboNN sheds light on how different mRNA sequences work in various tissues. So it enables the prospect of more targeted therapies. For conditions like cancer or inherited genetic diseases, treatments often falter due to insufficient protein production or ineffective targeting. RiboNN allows drug developers to construct mRNA therapeutics in a way that they express most in the tissue type of their choosing. And least in cells where it could cause damage. That could lead to more effective vaccines. Or treatments that are faster-acting, have fewer side effects and are cheaper to develop. Sanofi and UT Austin think the use of RiboNN could speed treatments for not only viruses. But also cancer and genetic diseases.
Challenges and Further Work with RiboNN
Despite its strong performance, RiboNN is also not a universal solution. The accuracy of the model relies heavily on the quality and diversity of its training data. Even if it’s trained across many cell types, unseen or hard-to-measure tissue-type conditions might still behave differently. Furthermore, the design of mRNA therapeutics is subject to a host of constraints beyond translation efficiency. Stability, delivery, immunogenicity and manufacturability are no less critical.
RiboNN provides one tool to optimise one important aspect, but downstream validation in the lab and clinical settings is essential. There’s also the issue of biological complexity — regulatory sequences, 3D structure, codon usage and cellular context could induce effects the AI model isn’t able to perfectly account for. Scientists working with RiboNN will still have to test their designs out in actual organisms.
To Conclude
RiboNN represents a significant leap toward smarter, more efficient mRNA therapeutics. By honing the accuracy of protein production prediction, particularly across different cell types, it holds the potential to shave development time, cut back on failed experiments and sharpen the precision of treatment design. Many challenges lie ahead. Such as delivery, immune effects, manufacturability, and real-world validation. But the arrival of RiboNN presages a future in which AI takes a central role in designing lifesaving treatments with speed and accuracy. As researchers adopt and refine this tool, it could very well become a linchpin of precision medicine — unlocking better treatments for cancer, genetic diseases, and much more beyond.
About The Author
