Quantum comes alive
Computational breakthroughs are making a difference in life sciences
Originally published March 15 2024
My father was born in The Netherlands in 1935, just a few short years before the start of World War II, which naturally affected his early education. During the war’s later years, his parents sent him from their home in Amsterdam to live on an uncle’s farm in the north. Dad’s formal schooling inevitably suffered but at the same time he learned a lot about agriculture and looking after livestock, especially cattle. When the war ended, he quickly made up for lost time and proved to be a very good student; by high school, it was obvious he was headed for post-secondary education. Indeed, he became the only one of his siblings to do so when he enrolled in the “Landbouwhogeschool”—the Agricultural College—in the town of Wageningen.
Dad took a liking to the academic life and to the town. Besides meeting the love of his life, marrying and starting a family, he stayed in university and went on to earn his PhD. In that, he didn’t stray far from his childhood on the farm, writing his dissertation on the use of blood typing to determine the possible parentage of a calf. This has applicability and often high value in maintaining purebred herds. After eventually leaving The Netherlands in 1967, Dad’s career centred on providing blood typing services for parentage testing to Canadian farmers. In 2002, shortly after his retirement, he was inducted into the Canadian Agricultural Hall of Fame.
As it happens, Dad’s old Agricultural College in Wageningen has become a European centre for quantum computing research. In fact, around the world, many universities, government agencies and private companies are investing heavily in quantum computing to accelerate the discovery process in virtually all aspects of life sciences. They’re using computer simulations to model scientific experiments, and the hope is that quantum computing can provide much more accurate simulations, much faster than classical computers can. They’re also using advanced mathematics to sort through huge experimental data sets in search of patterns and trends that will advance their research.
Let’s look at a few examples.
Gene theory
In 2023, I attended an IBM Quantum conference where I met a representative of the university in Wageningen, now known as Wageningen University Research (WUR). He told me of quantum research currently underway that I noticed builds, in a way, on Dad’s work from decades ago. The goal is to breed livestock with improved characteristics such as resistance to disease, higher meat or milk production in cattle or higher egg production in chickens, for example—and to do so as humanely as possible. Traditionally, scientists would select animals to breed based on observable traits and then wait a generation to see the results. Achieving measurable improvements could therefore take years. With advances in gene mapping, it’s now possible to take a mathematical approach to the selection process—which, it is hoped, will improve the odds of getting better results in the next generation.
This mathematical approach is called genomic prediction, part of the larger field of quantitative genetics. The idea is that for any given phenotypes, or physical characteristics, you can analyze the available population and select breeding candidates that will maximize the heritability of those phenotypes. This means analyzing millions of animals with thousands of genotypes, or DNA markers, which could affect the desired phenotypes. Genomic prediction involves setting up enormous systems of linear equations with potentially hundreds of millions of variables as the product of the population size, phenotypes and genotypes to be analyzed. The Wageningen researchers are currently using classical high-performance computing (HPC) to tackle the challenge but have recognized that the complexity of the problem is already exceeding the capacity of the best HPC available. They expect that quantum computers will be able to solve these systems of linear equations much faster. Faster and more accurate breeding candidate selection based on quantum computing will accelerate the overall breeding experiments, which in turn will yield higher and safer food production. The research is just starting, and preliminary results may be available toward the second half of 2024.
Quantum wellness
Resistance to disease is a desirable trait not just for livestock—it also underlies pretty much the whole human pharmaceutical industry. In late 2022, I attended an online symposium simulcast from Denmark and Germany that featured a presentation from the large pharmaceutical company Novo Nordisk describing challenges in the drug discovery process. Conventionally, discovery can take up to ten years and cost many tens of millions of dollars before the drug is ready for the market.
The process begins with the identification of a biological target—some physical manifestation in the human body related to a disease—along with new molecular compounds that may affect it. Qualifying these compounds requires extensive testing not only against the target but also for potentially negative interactions with other known drugs and other possible side effects. The possible solution space is staggeringly large. For approximately 19,000 prescription drugs currently approved globally, pharmaceutical companies estimate that the number of compounds first evaluated is in the hundreds of trillions.
The complexity only increases from there. The candidate compounds must be tested against each other to identify and eliminate side effects, and then tested for toxicity against all major organs. Finally, an appropriate delivery mechanism must be selected (e.g., spray, injection, tablet) which in turn may require testing with other compounds for coating, binding agents and the ability to manufacture the drug at scale. All this needs to happen before clinical trials on human subjects can even begin. If this testing is all done in vitro—literally in test tubes in the laboratory—it’s no wonder the scale of the time and financial investment is so large.
So, following a pattern we’ve seen before, scientists are turning to computational methods in an effort to simulate the molecular interactions, saving time and money and eliminating unproductive candidates prior to moving to lab research. This computational biochemistry requires modelling all of a molecule’s possible energy states, a task that grows exponentially with the number of electrons involved. On a classical computer, modelling a simple molecule like water (H2O) with three atoms and ten electrons can be done in seconds with only a few kilobytes of memory; moving to something slightly more complicated like methane (CH4) could take hours with a few gigabytes of memory. Once you get to the complex proteins used for pharmaceuticals, well, now it’s billions of years and the memory requirements are in the petabyte range (one petabyte is roughly equal to 500 billion pages of printed text).
The reason for this complexity goes back to quantum mechanics, which postulates that each electron in the system could be affected by all the other electrons, so that all possible pairs of electron-electron interactions must be considered. (Remember the travelling salesperson problem in optimization, where all paths between cities are evaluated? This is the biochemical equivalent.) There’s a lot more to it than that, but you get the idea.
Since quantum mechanics is at the root of the problem, it’s only natural that quantum computing could be part of the solution. Pharmaceutical researchers are hopeful that a quantum computer can intuitively do a better job of modelling the quantum states of molecules, thus reducing the memory and processing time requirements. IBM has estimated that a reasonably simple molecule like caffeine, consisting of 18 atoms, can be represented using 160 logical qubits. More complex protein molecules may take a few thousand logical qubits.
Logical qubits are expected to be assembled from large numbers—hundreds or even thousands—of physical qubits, to correct errors in large scale quantum computers. The technology is not yet far enough advanced, but some of the most promising areas of research include reducing the ratio of physical to logical qubits, using other error mitigation techniques and making qubits more robust and fault-tolerant. It won’t be long before quantum computers evolve to the level of complexity required for pharmaceutical molecular modelling—enabling faster and much less expensive discovery of safer, more effective medicines.
Mind matters
An especially welcome application of these breakthroughs will be in the field of mental health, which has in recent years finally been recognized as equally worthy of study, diagnosis and treatment as physical diseases have always been. The brain remains, probably, the most mysterious organ in the human body: the connection between the physical (neurons and synapses) and the psychological (human emotions and thoughts) still isn’t well understood. The problem, once again, is scale and complexity. Most estimates are that the brain is composed of approximately 100 billion neurons. Each neuron has direct connections to, on average, 10,000 other neurons, leading to an estimated 1,000 trillion synaptic connections between neurons.
Precisely mapping neural activity in different regions of the brain to observed human behaviours is already beyond the capability of classical computing. Combine this with the pharmaceutical discovery and development process for psychotherapeutic drugs and you might wonder how we’ve made any progress at all in this field. Although medicines do exist to manage conditions like depression, schizophrenia, psychoses and others, permanent cures are for the most part not yet available.
Quantum computing, again, can help. Mapping neural activity to human behaviour involves finding patterns in massive data sets—something that quantum optimization and quantum machine learning are particularly good at. Quantum neural networks (another connection of quantum with artificial intelligence) may eventually provide a better, more accurate model of the brain itself. Functional magnetic resonance imaging, or fMRI, is used to look at brain functions over time—and interpreting these images requires solving large sets of linear equations, another task well suited to quantum computers. And, as we’ve seen, quantum molecular modelling should improve the discovery of new psychotherapeutic medications.
Generational strides
There’s a lot more to life science than the three examples I’ve described above. But a safe and reliable food supply, affordable and effective medicine, and better understanding and treatment of mental health—these are among the biggest problems society faces today. If quantum computing can help with any or all of them, it will be worth the investment—because it means we will all be leading better lives.
It’s also gratifying for me to think that my father’s work, in some small way, was part of the foundation.