Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Scientists have succeeded shown a such as a computer it can improve the accuracy and reach of the output Artificial Intelligence Drug Discovery. And they did this by using their free time and the money left over for other projects.
A team at the Technical University of Denmark ran their AI model for protein prediction in conjunction with a quantum computer developed by British startup ORCA Computing, which accelerated AI by combining quantum machines with traditional processors. The researchers used a hybrid approach to create new peptides – short chains of amino acids – that can bind to specific proteins in the body. Doing so is an important part of vaccine development.
The team of researchers worked on weekends and collected funds that were not used from other sources because “the most new science is very scary at the base,” according to DTU professor Timothy Patrick Jenkins, who led the project.
Creating peptides in the laboratory and testing whether these would bind to specific proteins showed that the model produced better peptides than its former counterpart, with a powerful control where academic knowledge was lacking.
The team believes the system could speed up the development of immunotherapies and vaccines, and improve the effectiveness of drugs in underserved populations.
“We needed to make sure that the skeptics were convinced that our predictions fit the real world,” Patrick Jenkins tells WIRED. Quantum computing remains a a new garden and faced a lot of technical difficulties in building this machine and good use of problem solving.
Although Patrick Jenkins initially did not want to explore technology: “I was very skeptical” he says with a laugh, believing that any work would be “many years.”
He and his team use big data and AI to discover proteins that can unlock cheap and fast immunotherapies, often funded by the Novo Nordisk Foundation. While many biological modelers are eager to learn more, a particular problem for his group has been the lack of data on all types of human race, since most clinical research has focused on Western populations. This could make it difficult to develop peptides that would work in less educated populations, such as Asians and Africans, he says.
His team decided that putting more computers into their workflow could lead to different peptides, especially for targets with limited data, after learning that the machine had similar results in creating images.
The newly discovered method does not change research while quantum computers are still too small to run all kinds of AI, meaning that better results can be obtained on older computers.
“Quantum is still not very powerful, so the number of problems we can put into it is not predictable, which is what we usually work with,” says Jonathan Funk, a PhD student at DTU. Furthermore, finding a peptide that binds to a specific gene is only one step in vaccine development, and may not lead to a successful drug.
“I think it’s no surprise that many companies in the industry think that the scale of things is far from being seen,” ORCA Computing CEO Richard Murray tells WIRED, perhaps because the technology “has never had clear long-term models.”
He added that the research is new because it shows the long-term use of quantum computing. His company is also using technology through projects with oil major BP on chemistry and car manufacturer Toyota to improve its design.
The DTU team will now see if they can use the workflow with more expensive models and larger proteins. “We needed this as an easy way to prove that we now have the opportunity to move the needle more,” says Patrick Jenkins, noting that the spread of AI is especially important for neglected diseases that receive little research funding. He is also looking to use quantum computing to improve his approach to AI in manufacturing antidote for snake bites.
