UT-Sanofi Partnership Develops AI Model to Accelerate mRNA Therapeutic Discovery
Key Insights
Researchers from The University of Texas at Austin and Sanofi have developed RiboNN (search), an artificial intelligence model that predicts how efficiently specific mRNA (search) sequences will produce proteins (search) in different cell types.
The model demonstrated twice the accuracy of previous approaches when tested across more than 140 human and mouse cell types, potentially reducing trial-and-error experimentation in mRNA (search) therapeutic development.
RiboNN (search) could enable targeted therapies for specific organs like the liver or lung by optimizing mRNA (search) sequences for enhanced protein production in particular cell types.
A groundbreaking artificial intelligence model developed through a collaboration between The University of Texas at Austin and pharmaceutical giant Sanofi promises to revolutionize mRNA (search) therapeutic development by accurately predicting protein production from specific genetic sequences. The model, called RiboNN (search), addresses a critical bottleneck in developing next-generation mRNA vaccines and treatments for cancer (search), infectious diseases (search), and genetic disorders (search).
Enhanced Prediction Accuracy Transforms Drug Development
RiboNN (search) demonstrated remarkable performance improvements over existing methods, achieving approximately twice the accuracy of earlier approaches when tested across more than 140 human and mouse cell types. This enhanced precision could significantly reduce the trial-and-error experimentation that currently slows mRNA (search) therapeutic development.
"When we started this project over six years ago, there was no obvious application. We were curious whether cells coordinate which mRNAs they produce and how efficiently they are translated into proteins (search)," said Can Cenik, associate professor of molecular biosciences at UT Austin and co-researcher on the project. "That is the value of curiosity-driven research. It builds the foundation for advances like RiboNN (search), which only become possible much later."
The research team's findings were published in Nature Biotechnology, highlighting the model's potential to guide the design of new mRNA (search)-based therapeutics by predicting which sequences will yield the highest protein production or better target specific body regions.
Targeting Specific Cell Types for Precision Medicine
One of RiboNN (search)'s most promising applications lies in developing cell-type-specific therapies. The model's ability to predict translation efficiency across different cell types opens new possibilities for targeted treatments.
"Maybe you need a next-generation therapy to be made in the liver or the lung or in immune cells," Cenik explained. "This opens up an opportunity to change the mRNA (search) sequence to increase the production of that protein in that cell type."
This capability could prove particularly valuable for treating genetic disorders (search) where specific organs or tissues require therapeutic intervention, as well as for developing more precise cancer (search) treatments that target tumor cells while minimizing effects on healthy tissue.
Comprehensive Data Foundation Enables AI Breakthrough
The development of RiboNN (search) was made possible by the team's creation of an extensive training dataset compiled from over 10,000 publicly available experiments measuring mRNA (search) translation efficiency across different human and mouse cell types. This comprehensive data foundation allowed AI and machine learning experts from both UT and Sanofi to develop the predictive model.
The research was supported by funding from the National Institutes of Health, The Welch Foundation, and utilized the Lonestar6 supercomputer at UT's Texas Advanced Computing Center, demonstrating the computational resources required for such advanced modeling.
Coordinated Protein Production Reveals New Biology
In a companion paper published simultaneously in Nature Biotechnology, the research team made an additional significant discovery: mRNAs with related biological functions are translated into proteins (search) at similar levels across different cell types. While scientists have long understood that gene transcription into mRNAs is coordinated, this study provided the first evidence that mRNA (search) translation into proteins is also coordinated.
Open Access Tool Accelerates Global Research
The research team has made RiboNN (search) publicly available, enabling researchers worldwide to leverage the tool for their own investigations. Cenik expressed enthusiasm about the broader scientific impact this accessibility could generate.
"That's always wonderful, when you see other people leveraging [your work] in many creative ways that you never imagined," Cenik noted. "That's really where it has the biggest impact — when other people take it and use it in their research. That's super satisfying for me."
The model represents the culmination of six years of research, with AI implementation occurring primarily in the final three years of development. Despite incorporating cutting-edge artificial intelligence, Cenik views RiboNN (search) as a natural evolution of computational approaches he has employed for over two decades.
The successful academic-industrial partnership between UT Austin and Sanofi demonstrates how collaboration between universities and pharmaceutical companies can accelerate the translation of basic research into practical tools for drug discovery and development.