A (Quantum) Leap in the Dark?

Challenges in applying quantum computing to drug discovery

The foundations for quantum computers were laid in the early 1900s through several pioneering works in quantum mechanics, beginning with Max Planck’s introduction of quantised energy levels to explain black-body radiation. Decades later, in the late 1970s and early 1980s, physicists began exploring whether computation itself could be described by the laws of quantum mechanics. Paul Benioff showed that a theoretical computing machine could be formulated within quantum theory, while Richard Feynman argued that classical computers struggle to simulate quantum systems efficiently, and that quantum machines might do so more naturally.

Chemistry and materials science were seen from the outset as natural application areas for quantum computers because such machines could represent quantum mechanical behaviour directly, and the potential for quantum computers to have a transformative role in drug discovery has been mooted. Recent advances in quantum hardware have resulted in real devices that are accessible for use, turning what were once purely theoretical ideas into more concrete possibilities. However, virtually all the devices that are currently available are referred to as “noisy intermediate-scale quantum” (NISQ) computers. These systems operate with a relatively small number of qubits, the basic unit of information for a quantum computer corresponding to the classical computer bit. They have limited coherence times, the length of time for which the qubits exist in a combination of states rather than being limited to 0 or 1, and the systems display significant sensitivity to noise and error.

Whilst today’s quantum computers are powerful enough to run real experiments, and devices from multiple hardware platforms are available via cloud services, they are not yet capable of supporting the large, fully error‑corrected algorithms that are often discussed in longer‑term visions of quantum computing. Many of the most attractive quantum algorithms, particularly those targeting high‑accuracy molecular simulations, assume the use of fault-tolerant hardware that does not yet exist. In practice, quantum‑inspired classical algorithms or improved classical heuristics often deliver comparable results using today’s technology. When combined with the cost of running quantum processes, these limitations have important implications for the near-term impact of quantum computing on drug discovery. Indeed, many of the practical applications of quantum computers today rely on hybrid approaches, in which quantum processors are used for specific sub‑problems within larger workflows that are built on classical high‑performance computing (HPC). Understanding where quantum methods will genuinely add value remains a central challenge.

Quantum computers are not simply faster versions of existing machines, and quantum and classical computers excel at different types of problems. Many of the tasks central to drug discovery, which currently run on classical HPC and artificial intelligence (AI), will remain firmly in that domain. This reality reflects the diverse nature of drug discovery itself, where many areas are driven less by the faithful representation of quantum mechanical behaviour and more by data availability, statistical modeling, and large scale pattern recognition. In these areas, such as target identification, data integration, and decision-making across complex biological systems, classical HPC and machine learning approaches are already highly effective and continue to advance rapidly. This is also particularly true for later-stage discovery and early development activities, where population-level data, statistical modeling, and regulatory considerations play a dominant role.

However, for a small subset of problems in drug discovery, particularly those rooted in quantum mechanical behaviour, quantum approaches may offer something genuinely different. The key challenge is therefore not whether quantum computing will impact drug discovery in general, but rather how to identify the specific tasks and problem types where its distinctive capabilities can add the most value.

Quantum chemical calculations of molecular properties are an obvious area where quantum computing and drug discovery can intersect. Indeed, the underlying challenges of describing electronic structure, bond formation, and molecular energetics are inherently quantum mechanical. The accurate modelling of such properties remains one of the most computationally demanding tasks, and whilst classical methods have advanced considerably, particularly through clever approximations and increased computational power, many chemically relevant problems still scale poorly as the size of the system increases. Electronic structure calculations, which typically have a scaling between N3 and N6 (where N is the number of atoms in the system), quickly become computationally expensive for cases of practical interest, forcing trade-offs between accuracy and feasibility.

The limitations are especially apparent for problems such as the study of biochemical reaction mechanisms, strongly correlated electrons, or systems involving metal ions in high-multiplicity complexes. In such cases, classical approximations are simply inadequate. In principle, quantum approaches can address some of these challenges by representing molecular electronic states more directly. In practice, however, current applications are largely limited to small molecules or simplified models, often using hybrid quantum-classical algorithms that combine classical optimisation with quantum subroutines. The routine simulation of full protein-ligand systems, explicit solvent environments, or large biomolecular assemblies remains beyond the capabilities of today’s quantum computing tools.

While molecular simulation and quantum chemistry represent the clearest point of alignment between quantum computing and drug discovery, there are other areas of drug discovery where quantum computing approaches have been proposed. One such area is combinatorial optimisation and related search problems, where questions are framed as searches over very large spaces of discrete possibilities, for example exploring conformational space or selecting subsets of compounds from large chemical libraries. These problems are attractive from a quantum perspective because they often involve large combinatorial search spaces, which can be challenging for classical algorithms as the problem size grows. Quantum approaches, including annealing-based methods and approximate optimisation algorithms, appear to be attractive tools for navigating these spaces and finding (near)-optimal results

In practice, however, classical heuristics are already highly effective for many relevant optimisation problems, and quantum-inspired classical methods can deliver comparable performance to current quantum hardware. As a result, much of the work in this space is exploratory, focused on benchmarking, reformulating problems, and understanding where any genuine advantage might emerge, rather than on replacing the existing workflows.

Exploration around sampling and probabilistic modelling is another area of interest. Rather than predicting a single outcome, these approaches aim to explore distributions of states or configurations, which can be relevant to understanding properties such as molecular flexibility or energy landscapes. Here again, whilst there appear to be opportunities to use quantum computational methods, it is currently difficult to demonstrate practical advantages over classical approximations, particularly given current hardware limitations.

Quantum machine learning is another area that has attracted significant attention, driven in part by the success of classical machine learning across drug discovery. Hybrid quantum–classical learning models have been proposed for tasks such as pattern recognition or feature extraction, but many of the challenges that limit quantum computing more generally, such as noise, scale, and data overhead, also apply here. For data rich problems, classical AI approaches remain the dominant approach, and quantum machine learning remains an area of active research rather than practical deployment.

In the near term, quantum computing is likely to remain a largely exploratory tool in most areas, focused on benchmarking, algorithm development, and hybrid quantum-classical workflows that test where value might emerge. Over time, advances in hardware stability, scale, and error mitigation may enable more relevant applications, particularly for targeted molecular simulations where accuracy is a limiting factor. Longer term possibilities, such as fully fault tolerant quantum systems and the routine simulation of complex chemical and biological environments, appear more speculative and uncertain. Progress is unlikely to be linear, and any impact is expected to be incremental and selective rather than sudden or universal.

Nonetheless, many large pharmaceutical and biotech companies have been quick to forge alliances with specialist quantum computing firms in the hope of capitalizing on the technology’s potential. As far back as 2021, a review article noted that “Seventeen of the largest 21 pharmaceutical companies have publicly documented activities in quantum computing” and “Thirty-eight of about 260 quantum computing startups are tackling pharmaceutical problems”.  A more recent article at Fierce Biotech underlines this finding.

A pioneering example of a collaboration between a large biotech and two technology specialists is that between Biogen, Accenture Labs and 1Qubit. As far back as 2019, this partnership reported the successful development and application of a quantum-inspired method for 3D ligand-based virtual screening. A more recent (2021) agreement between Boehringer Ingelheim and Google Quantum AI includes the development of quantum methods for molecular dynamics simulations amongst its goals. In the same year, Roche partnered with Cambridge Quantum Computing (now Quantinuum) with a particular interest in applying quantum technologies in the field of Alzheimer’s disease research. Some other quantum computing companies are developing drug discovery platforms to offer to partners. These include POLARISqb’s QuADD and Qubit Pharmaceuticals’ Atlas. It will be interesting to see if these platforms can be shown to provide better results than well-attested classical analogues.

In summary, quantum computers are unlikely to completely replace the classical computational methods that underpin modern drug discovery. Instead, their potential lies in complementing existing approaches, with the biggest impact likely to be on challenging problems where classical methods struggle to balance accuracy and feasibility. As quantum hardware and the algorithms that run on it continue to mature, the key challenge is not to predict when impact will occur, but to make sure effort is focused on identifying the right applications of the technology at the right time in its evolution

If the future of quantum computing in drug discovery seems a bit uncertain, perhaps we shouldn’t be surprised. After all, uncertainty lies at the very heart of quantum mechanics!