Using artificial intelligence, MIT researchers have give you a brand new option to design nanoparticles that may more efficiently deliver RNA vaccines and other forms of RNA therapies.
After training a machine-learning model to research 1000’s of existing delivery particles, the researchers used it to predict latest materials that might work even higher. The model also enabled the researchers to discover particles that might work well in various kinds of cells, and to find ways to include latest forms of materials into the particles.
“What we did was apply machine-learning tools to assist speed up the identification of optimal ingredient mixtures in lipid nanoparticles to assist goal a distinct cell type or help incorporate different materials, much faster than previously was possible,” says Giovanni Traverso, an associate professor of mechanical engineering at MIT, a gastroenterologist at Brigham and Women’s Hospital, and the senior writer of the study.
This approach could dramatically speed the technique of developing latest RNA vaccines, in addition to therapies that might be used to treat obesity, diabetes, and other metabolic disorders, the researchers say.
Alvin Chan, a former MIT postdoc who’s now an assistant professor at Nanyang Technological University, and Ameya Kirtane, a former MIT postdoc who’s now an assistant professor on the University of Minnesota, are the lead authors of the brand new open-access study, which appears today in .
Particle predictions
RNA vaccines, similar to the vaccines for SARS-CoV-2, are frequently packaged in lipid nanoparticles (LNPs) for delivery. These particles protect mRNA from being broken down within the body and help it to enter cells once injected.
Creating particles that handle these jobs more efficiently could help researchers to develop even simpler vaccines. Higher delivery vehicles could also make it easier to develop mRNA therapies that encode genes for proteins that might help to treat quite a lot of diseases.
In 2024, Traverso’s lab launched a multiyear research program, funded by the U.S. Advanced Research Projects Agency for Health (ARPA-H), to develop latest ingestible devices that might achieve oral delivery of RNA treatments and vaccines.
“A part of what we’re attempting to do is develop ways of manufacturing more protein, for instance, for therapeutic applications. Maximizing the efficiency is very important to give you the chance to spice up how much we will have the cells produce,” Traverso says.
A typical LNP consists of 4 components — a cholesterol, a helper lipid, an ionizable lipid, and a lipid that’s attached to polyethylene glycol (PEG). Different variants of every of those components could be swapped in to create an enormous variety of possible mixtures. Changing up these formulations and testing every one individually could be very time-consuming, so Traverso, Chan, and their colleagues decided to show to artificial intelligence to assist speed up the method.
“Most AI models in drug discovery give attention to optimizing a single compound at a time, but that approach doesn’t work for lipid nanoparticles, that are fabricated from multiple interacting components,” Chan says. “To tackle this, we developed a brand new model called COMET, inspired by the identical transformer architecture that powers large language models like ChatGPT. Just as those models understand how words mix to form meaning, COMET learns how different chemical components come together in a nanoparticle to influence its properties — like how well it may possibly deliver RNA into cells.”
To generate training data for his or her machine-learning model, the researchers created a library of about 3,000 different LNP formulations. The team tested each of those 3,000 particles within the lab to see how efficiently they might deliver their payload to cells, then fed all of this data right into a machine-learning model.
After the model was trained, the researchers asked it to predict latest formulations that might work higher than existing LNPs. They tested those predictions by utilizing the brand new formulations to deliver mRNA encoding a fluorescent protein to mouse skin cells grown in a lab dish. They found that the LNPs predicted by the model did indeed work higher than the particles within the training data, and in some cases higher than LNP formulations which are used commercially.
Accelerated development
Once the researchers showed that the model could accurately predict particles that might efficiently deliver mRNA, they began asking additional questions. First, they wondered if they might train the model on nanoparticles that incorporate a fifth component: a variety of polymer often called branched poly beta amino esters (PBAEs).
Research by Traverso and his colleagues has shown that these polymers can effectively deliver nucleic acids on their very own, in order that they desired to explore whether adding them to LNPs could improve LNP performance. The MIT team created a set of about 300 LNPs that also include these polymers, which they used to coach the model. The resulting model could then predict additional formulations with PBAEs that might work higher.
Next, the researchers got down to train the model to make predictions about LNPs that might work best in various kinds of cells, including a variety of cell called Caco-2, which is derived from colorectal cancer cells. Again, the model was in a position to predict LNPs that might efficiently deliver mRNA to those cells.
Lastly, the researchers used the model to predict which LNPs could best withstand lyophilization — a freeze-drying process often used to increase the shelf-life of medicines.
“It is a tool that enables us to adapt it to an entire different set of questions and help speed up development. We did a big training set that went into the model, but you then can do way more focused experiments and get outputs which are helpful on very different sorts of questions,” Traverso says.
He and his colleagues at the moment are working on incorporating a few of these particles into potential treatments for diabetes and obesity, that are two of the first targets of the ARPA-H funded project. Therapeutics that might be delivered using this approach include GLP-1 mimics with similar effects to Ozempic.
This research was funded by the GO Nano Marble Center on the Koch Institute, the Karl van Tassel Profession Development Professorship, the MIT Department of Mechanical Engineering, Brigham and Women’s Hospital, and ARPA-H.
