IBM’s second-generation, 156-qubit Quantum Heron processors can run quantum circuits with up to 5,000 two-qubit gate operations using Qiskit. (Credit: Ryan Lavine for IBM)
In 2022, Moderna brought in approximately $19.26 billion in revenue, largely thanks to its groundbreaking Spikevax COVID-19 vaccine. In January 2025, the company is projecting revenue of $1.5 billion to $2.5 billion. To reverse this downturn, the company is pushing to broaden mRNA’s applications into cancer, rare diseases and other areas, but that requires cracking tough computational barriers in sequence design.
The therapeutic potential of mRNA extends far beyond COVID-19, or RSV, another infectious disease for which it scored an FDA approval in 2024. The ability to precisely instruct cells to produce specific proteins opens the door to a new class of medicines for a vast range of diseases. Yet designing the optimal mRNA sequence for a given therapeutic protein is a monumental computational challenge. Consider that the human body has more than 100,000 types of proteins, generated from about 20,000 genes through various modifications, and each protein is derived from mRNA. For any one of those proteins, the number of possible mRNA sequences creates a complex optimization problem that strains the limits of classical computation. The molecule’s secondary structure, the way it folds into stems, loops and bulges, compounds the problem, determining how efficiently the mRNA translates into protein, how stable it remains in the body, how it interacts with cellular machinery, and whether it triggers unwanted immune responses.
“Quantum computing lets us frame the mRNA-folding question as a giant puzzle in which every possible pattern of base-pairing is scored by its predicted free-energy,” says Wade Davis, Moderna’s senior vice president of digital. “It’s natural to look at quantum as another approach that could be complementary,” adds Sarah Sheldon, senior manager, quantum theory and capabilities at IBM. In its work with IBM, Moderna is applying this quantum-centric approach to the optimization of mRNA sequences. The aim is to expand the diversity of candidate molecules its design pipeline can produce for applications ranging from new vaccines to personalized cancer treatments.
The mRNA molecule naturally tends to adopt the structure with the lowest free energy, which is its most stable state. The central difficulty, as Davis puts it, is that “The computational bottleneck lies in searching for optimal solutions across an astronomically large design space.” While evaluating the quality of a single candidate sequence is relatively easy, finding the best ones is substantially more involved.
Cracking the computational hurdles in mRNA design
The mRNA folding problem is a good fit for quantum computers not because of “big data,” but because of its complexity. According to Sheldon, the ideal quantum problem has an “underlying structure that makes it hard classically at a relatively small size.”
The mRNA folding challenge exhibits exactly this trait. As the nucleotide sequence gets longer, the computational difficulty scales exponentially. Sheldon notes that while current work on 60-nucleotide sequences can be verified with classical computers, the problem “gets hard very quickly” beyond that point. This scaling challenge is what makes a quantum approach so promising for Moderna’s long-term goals. Davis emphasizes the future potential, stating the approach shows how maturing quantum devices could help scientists explore a “vastly broader landscape of inherently stable mRNA designs more quickly than classical methods alone.”
Quantum joins the biotech toolbox
Moderna’s strategy is one of early adoption, reflected in its move to hire dedicated quantum specialists, including professionals managing quantum algorithms and applications. This philosophy signals a trend where companies build in-house quantum skills. The trend underscores the growing need for hybrid experts who can connect industry-specific challenges to quantum algorithms.
The collaboration integrates quantum and classical systems into a single workflow. As Sheldon, explains, the process is about “figuring out where you need a quantum computer within your workflow,” using it for the specific computational bottlenecks that classical machines struggle with. This hybrid model is central to what IBM calls the “era of quantum utility.” For Moderna, the primary goal is building institutional knowledge, mastering the art of translating complex biological problems into a language quantum computers can solve.
This strategy creates a clear division of labor. While partners like IBM focus on developing better hardware, Moderna is mastering the application of that hardware to real problems. This will pay off in stages. In the near term, Davis explains, “the value is cutting search time on known targets.” The longer-term hope is that a “diversified sequence set could surface novel designs that classical heuristics may overlook.”
Inside the hybrid quantum process
Tackling the mRNA structure problem begins by translating it into a format a quantum computer can understand. The team maps the biological challenge onto a Quadratic Unconstrained Binary Optimization (QUBO) problem, a mathematical model used to represent complex decision problems as binary variables and quadratic objectives.
The method is inherently hybrid, using a Variational Quantum Eigensolver (VQE) that creates a feedback loop between quantum and classical machines. A quantum circuit is run, and the results are fed to a classical optimizer, which then adjusts the parameters for the next run. This iterative partnership is crucial, as “classical computers are very good at a lot of things,” Sheldon notes.
It’s really about figuring out where you need a quantum computer within your workflow. —Sheldon
To refine this process, the team incorporated Conditional Value at Risk (CVaR), a technique adapted from finance. Rather than averaging all possible outcomes from the quantum computer, CVaR focuses the algorithm on the most promising results. As Sheldon explains, it directs the search toward the “tail end of the distribution that has lower energies.” In other words, it represents the most stable molecular structures while allowing the optimization to converge faster.
This entire process yields a shortlist of promising nucleotide sequences. These are then rigorously validated. According to Davis, for smaller sequences, they perform “classical cross-checks… we confirm that the quantum routine reproduces the classical optimum.” For larger problems, they “compare the quantum-generated answer with the strongest classical results available,” ensuring the results are reliable.
Quantum’s potential impact on medicine
In 2024, the collaboration simulated mRNA sequences of up to 60 nucleotides on quantum hardware. Published in an arXiv preprint, the study demonstrated that a Conditional Value at Risk (CVaR)–based variational quantum algorithm could reliably reproduce minimum free energy secondary structures, matching the results of commercial classical solvers such as CPLEX. This scale is valuable because the results can still be verified against classical computers, providing a crucial benchmark. As the technology scales, Davis notes it could allow researchers to “treat ~100-nucleotide segments,” a size large enough to begin competing with state-of-the-art classical methods.
The ultimate goal is a quantum-enabled pipeline that could improve treatments and “make ultra-rare or even personalized mRNA treatments more accessible.” Yet Davis is quick to caution that these gains are a “hopeful research direction rather than an operational capability,” as they depend on future advances in hardware and error mitigation.
Realizing this vision requires overcoming several concrete hurdles. Davis identifies key challenges, including the scarcity of hardware, the skill gap requiring user-friendly tools for scientists, the need for regulatory traceability in an auditable format, and ensuring results can integrate cleanly with existing AI and molecular dynamics stacks.
Despite these obstacles, this work is creating a path forward. Sheldon sees the current era as a unique opportunity for innovation, where practical hardware allows researchers to rigorously test new ideas for both quantum algorithms and their real-world application.
Quantum’s reach beyond biotech
The current work is taking place in what Sheldon describes as a “pre-fault-tolerant world,” where today’s quantum processors rely on heuristic methods and error mitigation rather than full error correction. Despite these limitations, the collaboration is focused on foundational problems, quantum simulation and optimization, that have broad applications across industries like finance, logistics, and materials discovery.
Despite these limitations, there are broad cross-industry applications being actively explored. As Sheldon explains, “from my end, I look at it as, what are the algorithms that we’re developing, and what are the types of problems those algorithms can address?” Two major areas of focus are quantum simulation and optimization. There is a significant effort in “the simulation of quantum systems like chemistry and materials,” which has direct relevance for “drug discovery problems, but also materials discovery.” Additionally, “optimization problems pop up all over the place. Obviously finance, but also logistics. There’s a lot of hard optimization problems out there.”
Looking toward the future, IBM has a detailed quantum roadmap. This plan includes the development of the IBM Quantum Heron processor, which features lower error rates, and outlines a progression toward quantum-centric supercomputing that integrates with High-Performance Computing (HPC) resources. A key long-term goal of this roadmap is achieving eventual error correction capabilities.
Sheldon believes this synergy between hardware development and practical problem-solving is what defines the current innovation opportunity. “And I think that’s really what is exciting about the coming years,” she concludes, “is that we have hardware where we can really do this testing and understanding of the scaling of problems… there’s a lot of opportunity there for new ideas on the problem side and the algorithm side.”
The partnership is featured as a case study on IBM’s website.