Does Chinese tech giant’s AI algorithm hold the key to better mRNA vaccines?
- Baidu-led researchers say their method – using techniques from language processing – takes just minutes to find optimal sequence for more effective and stable Covid-19 jabs
- It can also be used to develop cancer medicine, monoclonal antibodies and treatments for other diseases, according to Nature paper
Their Covid-19 vaccine achieved up to 128 times the antibody response in mice compared to mRNA vaccines designed using conventional methods, and it could potentially remain stable at higher temperatures, eliminating the need for ultra-cold storage, according to an article published in the peer-reviewed journal Nature on Tuesday.
Vaccines developed with the algorithm may offer better protection at the same dose – or similar protection at a smaller dose with fewer side effects – compared to existing methods, the scientists said.
The researchers also applied the algorithm to create an mRNA vaccine for the varicella zoster virus to reduce the risk of shingles. They reported a six-fold increase in stability and an eight-fold increase in antibody response compared to vaccines developed with existing methods.
How China’s Covid-19 self-reliance medical goal could lose momentum
There are countless mRNA sequences that correspond to the coronavirus’s spike protein, and many of them are potentially stable at higher temperatures. But generating all possible sequences and testing them presents an “insurmountable computational challenge”, the scientists wrote.
To identify more stable sequences, the Baidu team borrowed techniques from natural language processing to develop their algorithm.
China approves first domestically made mRNA vaccine for Covid
In speech-to-text recognition, if a person says, “I like this meal”, it might sound like “alike this veal” or another combination of similar-sounding words. To build a sentence that makes the most sense among many alternatives, computational linguists use a technique that shows potential word sequences and picks the most plausible option based on grammar.
Just as sentences are made up of sequences of words with an underlying grammatical structure, RNA strands are built with sequences of nucleotides – the building blocks of nucleic acids – and a secondary structure based on its folding pattern.
Using the algorithm, the team created a graph representing all mRNA candidates, then applied the linguistic technique to identify the optimal mRNA sequence. It only took 11 minutes for the algorithm to find the best sequence – a task that would have taken billions of years using other methods, according to the scientists.
“We show that finding the optimal mRNA among the vast space of candidates is analogous to finding the most likely sentence among numerous similar-sounding alternatives,” the team wrote. “Our interdisciplinary approach is another example among the recent fruitful exchanges between linguistics and biology.”
04:48
Will ChatGPT replace reporters? We asked AI to write for the Post
Lead author Zhang He, a staff software engineer at Baidu Research, said the algorithm was a rare example of success in deploying AI in biomedical research.
“Without this AI algorithm, the region of highly stable sequences would remain inaccessible,” Zhang said. “We also built an unexpected bridge linking two fields of study that do not share obvious similarities at first glance.
“I hope that more AI technologies will be applied to biomedicine research and support human exploration in life science.”
The research team includes scientists from Baidu Research USA, Oregon State University, the University of Rochester Medical Centre, California-based start-up Coderna.ai, as well as Shanghai-based biotech start-up Stemirna Therapeutics.
In 2021, Baidu licensed the algorithm to French pharmaceutical company Sanofi for use in designing mRNA-based vaccines and medications. According to Reuters, Baidu will receive “milestone payments” whenever a product developed with the algorithm enters clinical trials.
