A team at the Technical University of Denmark says it has shown a quantum computer can make a generative AI drug-discovery model better at designing peptides, a narrow but useful step toward vaccines and immunotherapies. The work matters because peptide design is one of those biomedical chores where data gaps can quietly turn into worse tools for people underrepresented in medical research.
The group, led by DTU professor Timothy Patrick Jenkins, paired its protein-prediction model with a printer-sized quantum computer from British startup ORCA Computing. According to Jenkins, the researchers ran the project on weekends and used leftover money from other grants because, as he told WIRED, “most innovative science is too scary for foundations.” Science funding, apparently, still has a genre problem.
The system was used to generate new peptides, short chains of amino acids, intended to attach to particular proteins in the body. That binding step is one component of vaccine development. The researchers then made the peptides in a lab and tested whether they attached as predicted. According to the DTU team, the hybrid quantum-classical workflow produced more successful binders than a conventional version of the model, with the clearest gains in cases where training data was limited.
What the quantum machine changed
The claimed advantage was not that the ORCA machine replaced normal computing. The setup linked a quantum device with traditional processors inside an AI workflow. Jenkins’ team had suspected that adding a quantum system could push the model toward a more varied set of candidate peptides, especially for targets where the model had little data. They drew that idea from reports that quantum systems had produced similar diversity effects in image generation.
That matters in immunology because the data is lopsided. Jenkins told WIRED that his group, often funded by the Novo Nordisk Foundation, uses AI and large biological datasets to find proteins that could support cheaper and faster immunotherapy discovery. He said a persistent problem is that medical research has focused heavily on Western populations, leaving less genetic data for many people in Asia and Africa. A peptide model trained on that skewed record may have a harder time finding candidates relevant to understudied groups.
Jenkins said the lab work was needed because quantum computing still gets, and deserves, scrutiny. “We needed to really prove it to convince skeptics that our predictions connect to the real world,” he told WIRED. He also said he had been “a huge quantum skeptic” and had assumed useful applications for his own work were “decades away.”
Useful, limited, and not a drug
The result is not a shortcut to approved vaccines. DTU PhD student Jonathan Funk told WIRED that the quantum hardware is still too limited for the kind of full-size antibody work the team usually handles. “Quantum is still not very powerful,” Funk said, adding that the experiment could not encode the complexity of a normal-sized antibody.
There is another constraint: finding a peptide that binds to a target is only one stage in vaccine development. It does not prove safety, efficacy, manufacturability, or clinical value. Better results may still be possible on classical computers because today’s quantum machines remain small.
ORCA chief executive Richard Murray told WIRED that many industrial companies see quantum computing as distant and unclear because the field has lacked near-term examples of usefulness. He said this study is novel because it points to a commercial application that might arrive sooner. Murray also said ORCA is working with BP on chemistry projects and Toyota on design-process efficiency.
The DTU group now plans to test the workflow with newer models and larger proteins. Jenkins told WIRED he also wants to explore whether the method can help design synthetic antidotes for snakebite venom, and said generative AI tools may be valuable for neglected diseases that attract little research funding.
This story draws on original reporting from WIRED.
