Written by David Orrell
As the physicist Richard Feynman famously observed in 1982: “Nature isn’t classical, dammit, and if you want to make a simulation of nature, you’d better make it quantum mechanical.”
And as he went on, “by golly it’s a wonderful problem, because it doesn’t look so easy.” Indeed, the configuration of something like a drug compound, and its interaction with other substances, is ultimately a quantum problem which has so far eluded the power and reach of classical approaches. Enter quantum computers – which were only a dream when Feynman said those words, but now some 40 years later have become reality.
Quantum computers differ from their classical counterparts in that they are based on qubits, rather than bits. A classical bit can take on one of two values: 0 or 1. A qubit is a quantum system which exists in a superposed state – so a mix of 0 and 1 – until the time it is measured. A candidate storage device for a qubit would be a single electron, whose spin (a quantum variable analogous to axis of rotation) can be measured as either up or down; but there are many other options, such as photons, or charged atoms held in place by an electromagnetic field.
Quantum computers are enormously difficult and expensive to construct, because qubits are very sensitive to any kind of external perturbation or noise, and are susceptible to what physicists call “decoherence” due to interaction with the environment, which means that they lose their quantum nature. However their power scales up very quickly as more qubits are added. For example, one classical bit could represent one outcome (heads H or tails T) of a single coin toss. Four bits would therefore represent four particular outcomes in a row, such as HTHT. Four qubits however could represent all possible sequences (HHHH, HHHT, HHTH, etc.), of which there are sixteen, at the same time.
These states would exist in superposition, and only one would be measured at a time, so the system might have to be measured many times to get an accurate distribution. However, the quantum advantage becomes compelling as the number of qubits increases. A game with 50 coin tosses would have over a quadrillion possible outcomes, but could be represented by only 50 qubits.
At the moment, the most powerful quantum devices such as Google’s Sycamore chip (53 qubits) or IBM’s Eagle (127-qubits) are still fairly error-prone and therefore limited in their capabilities, however such devices are rapidly approaching the stage at which they will become useful for applications in science and business. Progress in the field is backed by billions in investments from giant technology companies, new startups, and government institutions (the Chinese state alone is said to have earmarked some $10 billion). Most people have never set eyes on a quantum computer, but companies including Amazon, IBM, Google, and Microsoft have already launched basic quantum-computing cloud services. And these firms expect to have fully functional quantum computers, which offer to significantly out-perform classical computers, available as soon as 2030.
Interest in areas such as engineering and finance is growing. A survey from the Certificate in Quantitative Finance Institute found that “when people were asked what areas they were most interested in, Quantum Finance and Quantum Computing were the most popular, after Machine Learning.” So what does this all mean for the pharmaceutical industry?
Drug development is a famously slow, risk-prone, and expensive business, with new drugs typically costing a couple of billion dollars in R&D, and taking ten years or more to reach market. One way that quantum computing can help this process, and potentially reduce costs, is by allowing researchers to directly simulate compounds, just as Feynman had proposed. At the moment, computer-assisted drug discovery (CADD) can handle only the simplest molecules, and even then relies on numerous approximations, which is one reason why so much of drug development depends on trial and error. In principle, a quantum computer can simulate the properties of a molecule directly, by using one quantum system to model another.
Techniques such as machine learning are currently limited by computer power. For example Google’s DeepMind developed AlphaFold to solve protein folding using deep learning, but takes weeks to perform its computations. Quantum AI, once it gets up and running, will be far more powerful.
In general, it is clear that quantum computing will speed up any problem currently tackled using conventional computers in areas including target detection, disease diagnosis, toxicity analysis, logistics, literature research, and so on – while also opening up entirely new applications.
Given that drug development is a $1.5 trillion industry, McKinsey estimated that an improvement of 1 to 5 percent would result in $15 to $75 billion in additional revenue. As they conclude, “Leaders should start to formulate their quantum-computing strategies, especially in industries, such as pharmaceuticals, that may reap the early benefits of commercial quantum computing.” According to Yianni Gamvros, head of business development at the Palo Alto-based quantum-software company QC Ware, “We expect the market to take off in two to three years for pharma and materials applications.” The latest entrant in this space is Quantinuum, formed in 2021 by Honeywell Quantum Solutions in partnership with Cambridge Quantum Computing, who list “Drug discovery and delivery” as one of their top use cases.
Fully exploiting the quantum advantage will take a deep pool of both money and expertise. However it is becoming clear that leaders in the pharmaceutical industry need to develop a quantum strategy in order to navigate what looks like being an increasingly quantum future.