Scientists have succeeded shown a such as computers can improve accuracy and generative access examples of the invention of artificial intelligence drugs. And they did it using their spare time and money left over from other projects.
The Technical University of Denmark team ran their generative AI model for protein prediction in conjunction with a printer-sized computer built by UK company ORCA Computing, which accelerated AI by combining quantum machines with traditional processors. The researchers used a hybrid approach to produce novel peptides—short chains of amino acids—that can bind to specific proteins in the body. Doing so is an important step in vaccine development.
The research team worked on weekends and pooled unused funds from other projects because “a lot of innovative science is fundamentally scary,” according to DTU professor Timothy Patrick Jenkins, who led the project.
Making peptides in the lab and testing whether these would bind to certain proteins showed the model produced more successful peptides than the old one, with significant improvements where training data was sparse.
The team believes that the machine could speed up the development of immunotherapy and personalized vaccines, as well as improve the effectiveness of drugs in underserved groups.
“We needed to prove it to convince critics that our predictions connect to the real world,” Patrick Jenkins tells WIRED. Quantum computing is still a young field and is facing intense scrutiny due to the technical challenges of building these machines and use them successfully to solve problems.
Even Patrick Jenkins was initially reluctant to explore the technology: “I was very skeptical” he says with a laugh, believing any use of his work would be “decades away.”
He and his team use big data and AI to discover proteins that could unlock new immunotherapy cheaply and quickly, often funded by the Novo Nordisk Foundation. Although many creators of biological models desire more data, a particular challenge for his team has been the lack of data on the exact types of genetic information in the human population, since most medical research has focused on Western populations. This can make it difficult to develop peptides that will work in uneducated populations, such as those in Asia and Africa, he says.
His team hypothesized that embedding a quantum computer into their workflow could enable them to generate more diverse sets of peptides, especially for targets where they had less data, after learning that the machines were equally effective at generating images.
The newly discovered process won’t revolutionize research yet as quantum computers are still too small to run full-scale, sophisticated AI models, meaning much better results can be achieved on a regular computer.
“Quantum is still very weak, so the level of complexity we could encode was not a normal-sized antibody, which is what we usually work with,” says DTU PhD student Jonathan Funk. Furthermore, finding a peptide that can bind to a specific gene is only one step in vaccine development, and not the only one to produce successful drugs.
“I think it’s not surprising that a lot of industrial companies think that population is bad and far away,” ORCA Computing chief executive officer Richard Murray tells WIRED, because the technology “hasn’t had clear close examples of benefits.”
He says this research is novel in that it shows the near commercial use of quantum. His company is also using the technology through projects with BP’s oil chemistry major and Toyota to make its design process more efficient.
The DTU team will now see if they can use the functionality with more sophisticated structures and larger proteins. “We needed this as an easy way to prove that we now have a chance to move the needle in a big way,” says Patrick Jenkins, noting that AI workflows are especially important in neglected diseases that receive little research funding. He is also looking at using quantum computing to improve his method of generating AI by design Synthetic antidote for snakebite.
