The heads of digital research technology at the top drug companies run into one another at conferences, and they frequently end up sitting side by side on panels. Discussions are simultaneously high level and down in the weeds, touching on trends in the increasingly data-dependent world of drug discovery. Recently, conferences and panels have focused on something so cutting edge that it isn’t even there yet—quantum computing in drug discovery.
It is a topic on which they want to gather more frequently.
“The idea came about at the Quantum.Tech conference in Boston in September of 2019,” says Brian Martin, head of artificial intelligence in R&D information research at AbbVie. “We were standing in a circle talking, and I realized we had representatives from GSK, Takeda, Pfizer, Merck, myself from AbbVie.”
It was a powerful circle, made up of people at the highest level of analytical and computational experience. “Emir Roach from Takeda said maybe we should formalize and start working together on this because none of us on our own is going to be able to make the case to our executives to spend money on this,” Martin recalls.
The group met 3 months later at the Q2B conference in San Jose, California, for the first time as a consortium called QuPharm. It has since begun pooling expertise and ideas about potentially beneficial applications of quantum computing—an ultrafast, high-volume discipline that is making strides at IBM, Google, and other technology firms but has yet to find a practical application in drug discovery.
Martin, Roach, and their peers foresee a major role for quantum computing in reducing time and improving results for in silico drug discovery—the growing practice of looking for new molecules with computers rather than test tubes. Research information technology heads are eager to advance as far as possible in precompetitive joint efforts so they can be ready as soon as quantum computing gains traction in the pharmaceutical research lab.
“We aren’t talking about an AbbVie compound,” Martin says. “Emir isn’t talking about something in Takeda’s pipeline and how quantum computing is going to help that specific compound. We are talking about general approaches towards things like simulation. We are talking about general problems that everyone understands, that we all have as companies working in the chemical and pharmaceutical space.”
QuPharm is collaborating with the Quantum Economic Development Consortium (QED-C), a group launched in 2018 with funding from the US National Institute of Standards and Technology. QED-C enables efforts to develop commercial applications for quantum information science and technology. QuPharm is also working with the Pistoia Alliance, a consortium of pharmaceutical information technology managers covering laboratory computation broadly.
The primary challenge to implementing a scalable quantum computing system—one that can handle parallel runs of computation on increasingly complex problems—is identifying how such a system will fit into the rhythm of the lab.
“For most problems in computational chemistry for drug development, classical computing is sufficient,” Roach says. “There are situations where there are limitations. What we haven’t been able to do now as an industry is pinpoint among those computationally limited problems which ones are amenable to resolution by quantum computing.”
Another challenge is translating chemistry problems into quantum computational problems. And labs lack proven hardware and software for a system that would demonstrate the benefits of the technique. Finally, there is the problem of who will do any of this work.
“We don’t have the talent to run the elements of scalable quantum computing in pharmaceuticals in a sustainable or repeatable fashion,” says Roach, a medical doctor by training who studied engineering and worked as a consultant to the drug industry at McKinsey & Company before joining Takeda Pharmaceutical as the head of emerging technology last year.
Others in similar positions at drug companies are importing know-how from seemingly unrelated sectors. Martin, for example, worked in finance, where he was one of the architects of the computer system developed to manage mortgage-backed securities after the 2008 stock market crash.
Martin says there are enough data in the public domain for QuPharm to make preliminary headway. He says the group has already compiled a list of 24 possible uses of quantum computing in the research lab, including such drug development staples as target identification, lead discovery, lead optimization, and clinical development. If QuPharm members have an actual target or molecule they would like to get started on, they could break off and form a partnership with a technology provider.
“Nothing has gotten to that stage yet,” Martin says. “There are multiple engagements happening in the pharma industry with quantum computing vendors exploring what can be done in a particular space, but the purpose of those is generally going to be much more educational and informative in nature.”
An example is Biogen, which partnered with the quantum computing software supplier 1QBit and the consulting firm Accenture to develop a quantum-enabled molecular comparison tool.
The project began with work Accenture was doing with 1QBit to identify drug discovery uses for a graph-based optimization algorithm under development at the software firm. The two approached Biogen, an Accenture client, with a proposal to use the technique to review molecule matches, predict positive therapeutic effects, and assess side effects in early-stage discovery.
Carl Dukatz, principal director in Accenture’s technology incubation group, explains that 1QBit employs what are called quantum-inspired methods that run on both quantum and classical computers; access to quantum computers happens via the cloud. The technique uses mathematical formulas to convert molecular structures into graphs for evaluation, allowing more precise comparisons than digital techniques provide, he says.
In an experiment, Biogen applied the approach to virtual screening of large molecular databases, says Govinda Bhisetti, the firm’s head of computational chemistry (J. Chem. Inf. Model. 2019, DOI: 10.1021/acs.jcim.9b00195). “We found it relevant to our in silico drug discovery work and recognized its potential to speed up drug discovery.”
Indeed, the quantum computing approach performed as well as or better than classical computational methods, Bhisetti says. “We have also found novel hits which were not found by classical methods.”
Biogen has not yet implemented the quantum-like computing operation in practice, Bhisetti acknowledges, but the company is continuing its collaboration with 1QBit to find effective uses of the technology.
Bhisetti is also involved in QuPharm. He says the pooling of know-how across the major drug companies has been crucial to grasping the range of chemistry problems that can be solved at the quantum level—“big problems that involve too many combinations of variables for classical computing to solve,” he says.
Another industry project links the German company Merck KGaA with Seeqc, a start-up that is designing quantum computing hardware in which the controls are on each individual chip. The firm aims to reduce the cost of running a quantum computer largely by reducing the heat load on the processor. Merck, which will access Seeqc computers via the cloud, has invested $5 million in the start-up through its venture capital arm.
Philipp Harbach, head of in silico research at Merck, notes that scientists have applied quantum mechanics to chemistry for over 20 years; after all, quantum systems describe every chemical reaction or property of a material. And quantum mechanics are applied in the same way as traditional chemical informatics techniques—with classical computers.
“Quantum computing will never replace classical computing” in drug discovery, Harbach says. “It will augment the existing classical computing technology.” He notes that quantum computing software start-ups like 1QBit are generally taking the augmentation approach as they develop cloud-based quantum algorithms and applications.
Alyn Chad Edwards, product manager at the software developer Cambridge Quantum Computing, describes a volatile dynamic between new and old techniques. “We are approaching the upper bounds of what is possible using classical machines. Even the world’s largest supercomputers struggle to solve chemistry problems,” he says. “On the flip side, we are at the lower limits of quantum computing. Whilst they are misaligned in terms of readiness, what we are starting to see is some cross-pollination between the classical and quantum worlds.”
As quantum computing advances as a discipline, quantum algorithms will be imported into classical computing systems in an evolution that will take years. “We will not all of a sudden switch gears and, bang, we will have quantum computing,” Edwards says.
Christopher Savoie, CEO of Zapata Computing, a service firm whose lead product is designed to coordinate classical and quantum computing, agrees. “A lot of this is future-oriented stuff,” he says. “Quantum-inspired algorithms running on classical machines won’t handle the level of complexity that an eventual quantum computer will be able to, but you can start getting stuff done now. You don’t have to sit on the sidelines waiting 5 years for that perfect quantum computer to show up.”
Managers at major drug companies, even those in the executive suite, want to be ready, Savoie says. “I think there is a realization that is beginning to emerge in the C-suite.” Top management is looking beyond the usual questions, such as return on investment, used to decide whether to commit to new technology, because it recognizes the potential for early adopters to gain critical advantage in R&D.
The C-suite may also see quantum optimization as a means of improving supply chain management. “There is an emerging recognition that this technology is going to be disruptive across all the digitalization efforts at these companies,” Savoie says.
But the talent question looms. “You can’t just take your data analytics people in the back room and convert them into quantum computer programmers overnight,” Savoie says. Companies have to consider workforce development now to be ready for a new level of computing that may not exist in practical application for several years.
Celia Merzbacher, deputy director of QED-C, says the formation of QuPharm shows the drug industry is looking ahead and preparing for the changes likely to be wrought by quantum computing. “By banding together, they are likely to create a significant market pull that accelerates the progress of hardware and software development that meets their needs,” she says. “I give them credit for being early to the table, for thinking about it, asking the questions, and paying attention.”
While many in quantum computing see small-molecule drug discovery as a likely front line in deploying the technique, Merzbacher says QED-C and software and hardware developers will be watching the drug industry closely to see what it actually does with the emerging technology.
They may be watching for quite some time, Merzbacher says, but that’s a good thing. “The pharmaceutical industry is patient,” she says, pointing to the decade or longer that it takes to advance a drug candidate from the lab to the pharmacy.
“That sort of culture is, I think, very helpful. They are not going to be the ones who develop the tools, but they have such a long pipeline that they can envision where quantum computing might intersect that pipeline in the future perhaps a little better than other sectors.”