A new AI-driven software has written code for COVID-19 vaccines that trigger up to 128 times greater antibody responses.
The research arm of Chinese tech giant Baidu has unveiled a novel algorithm to design mRNA vaccine sequences that were previously out of reach.
In validation tests with mice, the AI-driven software yielded code for COVID-19 vaccines that trigger up to 128 times greater antibody responses.
The algorithm, called LinearDesign, has also proven successful in extending the shelf stability of the vaccines sixfold, even when exposed to body temperatures.
Cold storage is needed for most vaccines, and this condition makes distribution efforts challenging, especially in the hotter regions of the world.
The ability to extend the shelf stability of vaccines and improve their tolerance to various temperatures could therefore have a radical impact on immunisation efforts across the planet.
The AI tool the researchers developed could also have applications beyond vaccines and help design potent new cancer treatments, Dr He Zhang, Staff Software Engineer at Baidu Research, told Euronews Next.
Just last week, personalised mRNA vaccines for pancreatic cancer patients showed promising results in a small study conducted by New York researchers and Germany’s BioNTech, staving off the return of the tumour in half of those treated.
The study used a pancreatic cancer mRNA vaccine tailored to each patient’s tumour to potentially help provoke an immune response.
Messenger RNA (mRNA) medicine refers to a new class of drugs and vaccines that use a small piece of genetic material called mRNA to teach the body's cells to produce a protein, which triggers an immune response against a specific pathogen, such as a virus. This approach is different from traditional vaccines, which use weakened or inactivated parts of a specific pathogen to stimulate the immune system.
Oncology is an important area for mRNA vaccine manufacturers, allergies is another. BioNTech - the same company that along with Pfizer produced the mRNA COVID-19 vaccine - also holds the patent rights to an mRNA vaccination platform designed to protect against allergens such as grass pollen and house dust mites.
While mRNA medicine is promising, it comes with its own challenges.
“mRNA vaccines have saved many lives,” Zhang said, “but there are still some problems around the stability and therefore the effectiveness”.
The challenge of keeping mRNA intact
Messenger RNA is unstable because it is single-stranded, unlike DNA, which is double-stranded.
The single-stranded parts of our mRNA are more easily cut down to pieces or degraded into segments - either by the immune systems, water molecules or even before being injected into your body, explained Dr Liang Huang, Professor at Oregon State University and co-author of the LinearDesign paper.
And when the mRNA is degraded into little pieces, it cannot pass the full message on to the cells.
“The goal is to keep the mRNA intact in its full length,” said Huang, and to do that the messenger RNA has to be as compact as possible.
“In other words, you want to design mRNA that folds and looks more like a [double-stranded] DNA,” he told Euronews Next.
And that’s exactly what the LinearDesign algorithm has done.
How does the LinearDesign algorithm work?
The breakthrough algorithm that promises to revolutionise the field of mRNA therapeutics is not actually new.
The LinearDesign model is based on a 1961 algorithm called lattice parsing, originally invented for natural language processing and speech recognition.
“We basically used the same algorithm without any change, the only change is the input, which was grammar,” he said.
The lattice parsing algorithm had been originally used to identify the most likely sentence among several possible alternatives that sound similar.
“The technique was already somehow there 60 years ago,” said Huang. “It was just waiting for us to discover the connection between two distant, remote fields: linguistics and biology”.
Using this technique, the researchers found a way to make the algorithm create, in just 11 minutes, “the most stable COVID-19 mRNA vaccine”.
In other words: they identified the most secure structure for the mRNA messenger to travel and deliver the message to the cells.
Running all the possible combinations that the AI algorithm did to find the most stable vaccine would have taken humans "the life of the universe: billions of billions of billions of years,” said Huang.
Is AI the future of vaccines?
The Baidu researchers predict the LinearDesign algorithm will make mRNA technology even “more popular than it currently is,” as well as enable other variants of the technology to enter more fields of medicine.
French giant Sanofi, a pharmaceutical industry company headquartered in Paris, has already licensed the LinearDesign technology. The drugmaker picked it up at the end of 2021 when the researchers first published the preprint of their study.
The paper - written in collaboration with Oregon State University, StemiRNA Therapeutics, and the University of Rochester Medical Center and unveiled in the scientific magazine Nature earlier this month - was first published under “rare option of accelerated article preview,” said Huang, which means the scientific journal made the study public before it actually underwent the editing process.
“The editor chose us for accelerated article preview because he, quote: ‘thought this has relevance to the public health and the pandemic’”.
The Baidu Research scientists are optimistic that mRNA technology will replace traditional vaccines in the next decade.
“I think in about five years, probably all the mRNA pharmaceutical industry will be using this algorithm,” Huang told Euronews Next, adding that “mRNA vaccines are much better, except for stability, which we solved”.
[Editor's note: This article has been corrected to clarify that the company behind the study is Baidu, China's largest search engine. Baidu Research is the research unit of Baidu.]