Bayer drugs teams up with Google Cloud to usher in a new era of drug discovery.
Drug discovery can be a herculean task. Despite many recent advances in understanding disease, the real challenge remains turning those insights into new drug treatments.
Consider the ongoing COVID-19 pandemic. Within a few months of the illness emerging, scientists had isolated the virus, determined how it spread, and learned how it infects our cells to make us sick. It would take another year until a vaccine was publicly available, a feat that took an incredible amount of effort, money and global cooperation.
For diseases of lesser global impact, the process can take decades. It’s one of the main reasons pharmaceutical companies continue to spend about 15 percent of their total sales on R&D. Developing new drugs remains slow, tedious and expensive.
In an effort to transform the traditional drug discovery pipeline, Bayer recently announced a new partnership with Google Cloud to drive their large-scale quantum chemistry operations. The company’s goal is to accelerate the discovery of drug candidates using quantum computing strategies.
In their joint press release, the companies expressed their desire to demonstrate that quantum-based simulations can reshape how we develop drugs and get treatments to patients faster than ever before.
So, How Do We Discover Drugs?
Depending on how you choose to analyze the history of medicine, drug discovery can date back to the beginning of civilization, when early humans discovered the healing properties of certain plants and animals. Expanding on thousands of years of traditional knowledge, many of our earliest medications were identified as natural products produced by plants, animals or microbes.
One clear example of the not-so-distant past is the common antimalarial drug chloroquine. Indigenous peoples in South America determined that tea made from the bark of the Cinchona tree could treat chills and fever, common symptoms of malaria. In the early 1800s, scientists determined that quinine, the first antimalarial drug, was the source of the activity in the tree’s bark. Later, scientists at Bayer discovered a synthetic analog of quinine, chloroquine, with potent antimalarial activity. It remained the leading antimalarial drug for decades and saved innumerable lives globally.
Unfortunately, many of these successful drug discovery stories stretch decades, and companies have long sought ways to accelerate their development pipelines. In recent years, this has involved high-throughput screening techniques, where researchers screen large collections of molecules for an activity of interest. For example, researchers have used such screens to identify candidate drugs for treating bacterial and fungal infections and even cancer. Although often faster than traditional natural product discovery, these screens rely on a limited resource: collections of small molecules.
Today, computational biology and chemistry now help researchers to accelerate the early stages of drug discovery. As long as a researcher has an identified biological target, computational methods can be used to identify molecules that could inhibit that target and treat its associated disease. Usually, this involves researchers modeling how well the structure of their target fits a given molecule, just like two puzzle pieces. The problem is the sheer computational power required to test millions or billions of molecules against a single target structure. Each simulation also has intrinsic uncertainties based on the principles of quantum mechanics and how the atoms of a molecule are arranged, further complicating the process.
So, researchers need a strategy to conduct billions of simulations across any number of biological targets of interest, with each pairing relying on complex quantum uncertainties.
A Quantum Solution
The principles of quantum chemistry are a direct extension of quantum mechanics. Based on decades of research, the physics of how molecules behave and how enzymes are structured is relatively well-defined. Essentially, every molecule is actually a quantum system, and quantum chemistry is used to approximate the location and arrangement of atoms. The issue is that these calculations rely on uncertainties, and exact methods are not possible with classical computing strategies. Instead, most companies have used approximations that are often plagued by inaccuracy. Some researchers have therefore speculated that quantum computing might be best utilized for simulating the quantum uncertainties of these drug-target pairings.
Companies can take advantage of large-scale machine learning algorithms that can utilize the principles of quantum mechanics. This can lead to the full adoption of quantum computing in the near future, either for partial or complete solutions to quantum chemistry-related problems.
In a June 2021 article on quantum computing in drug research, McKinsey & Company highlighted the potential for quantum computing to improve strategies scientists already use for initial drug discovery. For example, many companies already model the interaction of their target of interest with digital collections of small molecules. But with classical computing strategies, the structure of the target must be defined and largely inflexible. Although this makes the simulations possible, it does not truly reflect the biology, where protein targets are flexible and can often adopt any number of structures. Quantum computing would allow companies to integrate this biologically relevant flexibility into the screening, helping to identify and prioritize leading drug candidates.
Bayer Partners with Google Cloud
Bayer says it will use Google Cloud’s computational prowess to demonstrate that large-scale quantum mechanics can be employed to streamline computational chemistry. It will use Google Cloud’s (Tensor Processing Units) TPUs to develop complete quantum mechanical modeling of protein interactions with candidate molecules. Together, the two companies aim to determine if these calculations, at scale, are possible and economically feasible for applications in drug development.
“Accelerating drug discovery may be one of the most important applications for AI and high performance computing in the healthcare industry,” said Thomas Kurian, CEO of Google Cloud. “Bringing Bayer’s powerful research and development capabilities together with our industry-leading infrastructure has the potential to unlock new discoveries—with greater accuracy and speed—helping to get medicines to patients faster.”
It stands to reason pharmaceutical companies may be specifically interested in Google Cloud after witnessing the recent outputs of DeepMind, another subsidiary of Alphabet and Google Cloud partner. In 2021, they released the ground-breaking AlphaFold 2.0, capable of predicting the structure of proteins with unprecedented accuracy. The combination of AlphaFold and AI-driven drug discovery could prove powerful, especially for drug targets of interest for which researchers have not yet determined the structure.
Bayer’s announcement is not the first within the pharmaceutical industry. In 2017, the pharma company Biogen announced their partnership with Accenture Labs to utilize quantum computing for large-scale drug discovery. In collaboration with quantum software company 1QBit, the companies developed a quantum-enabled molecular comparison solution to improve the identification of drug candidates for diseases like Alzheimer’s and ALS.
There are also startups emerging with the expressed purpose of applying quantum computing to solve the biggest problems facing the pharmaceutical industry. Algorithmiq, based in Finland, and Qubit Pharmaceuticals, based in France, are just two examples of companies looking to lead in this space. Other startups are also already applying AI to drug discovery, including UK-based BenevolentAI.
An Industry Hungry to Embrace the Future
The pharmaceutical industry has long relied heavily on R&D to drive their business success. But with so much riding on R&D, pharma companies perpetually seek to streamline and optimize their drug discovery pipeline. AI is already being adopted across industries to improve the efficiency of manufacturing and large-scale company operations. But Bayer and other pharmaceutical companies are now looking to integrate AI directly into the heart of their operations. By incorporating quantum principles into early-stage drug discovery, pharmaceutical companies will be able to improve the accuracy, and hopefully the potency, of their initial candidates. Over the next few years, it will be exciting to enter the quantum computing renaissance and hopefully see the results of these partnerships emerging across the pharma industry. Only time will tell if the widespread excitement for quantum-enabled drug discovery will lead to faster, better drugs in the clinic.
