Messenger RNA (mRNA) carries the instructions for creating specific proteins, playing a vital role in supporting growth and essential biological functions. In modern medicine one of the most promising frontiers is the development of mRNA-based vaccines and therapeutics to combat viruses, cancers, and genetic diseases. However, a major challenge lies in ensuring that a patient's cells can produce enough protein from the therapeutic mRNA to effectively treat the disease.

A new model, named RiboNN aims to streamline the development of mRNA therapeutics by predicting which designs will yield the highest protein production or target specific organs like the heart or liver. The researchers introduced the model in one of two companion papers published in Nature Biotechnology.

“When we began this project more than six years ago, there wasn’t a clear application in mind,” said Can Cenik associate professor of molecular biosciences at UT Austin, who co-led the study with Vikram Agarwal, head of mRNA platform design data science at Sanofi’s mRNA Center of Excellence. “Our curiosity about how cells regulate mRNA production and translation efficiency laid the groundwork. 

This research was made possible with funding from the National Institutes of Health, The Welch Foundation and access to the Lonestar6 supercomputer at UT’s Texas Advanced Computing Center.

When tested across more than 140 human and mouse cell types, RiboNN demonstrated roughly twice the accuracy of previous methods in predicting translation efficiency. This significant improvement could empower researchers to make more precise predictions across various cell types—potentially accelerating the development of treatments for cancer, infectious diseases, and genetic disorders.

To understand how cells produce proteins imagine a team of chefs baking cakes. These chefs—your cell’s ribosomes—consult a recipe from a unique cookbook (your DNA), transcribe it onto notecards (messenger RNAs or mRNAs), and then use the listed ingredients (amino acids) to “bake” proteins.

mRNA vaccines or therapies work by guiding these cellular chefs to produce specific proteins. In vaccines, for example, they might prompt cells to make a viral or cancer-related protein, acting as a red flag for the immune system to generate antibodies. In the case of a genetic disorder, the mRNA might encode a protein the body cannot make on its own, helping to correct the deficiency.

building the RiBefore boNN model Cenik and the UT team assembled a large training dataset from over 10,000 publicly available experiments. These experiments measured how efficiently different mRNAs were translated into proteins across a wide range of human and mouse cell types.

One aim of the RiboNN tool is to enable cell-type-specific therapies, said Can Cenik, a CPRIT scholar and affiliate faculty at UT’s Oden Institute. “Whether the therapy needs to work in the liver, lungs, or immune cells, this tool can help tweak mRNA sequences to boost protein production where it’s needed,” he explained.

In a companion Nature Biotechnology paper, the team found that mRNAs with similar biological roles are translated at consistent levels across cell types—a newly discovered layer of coordination beyond gene transcription.

UT undergraduates helped build the training dataset RiboBase by verifying and completing public data. The RiboNN project was a collaboration between UT and Sanofi, with contributions from UT grad student Logan Persyn and Sanofi scientists Dinghai Zheng and Jun Wang supported by a partnership coordinated through UT’s Discovery to Impact office.

Source:https://news.utexas.edu/2025/07/25/new-ai-tool-accelerates-mrna-based-treatments-for-viruses-cancers-genetic-disorders