To solve the greatest challenges, it often makes sense to turn to the most advanced technology. In healthcare, the present challenge has never been greater: how to combat the coronavirus pandemic. Unsurprisingly, cutting-edge technologies such as artificial intelligence (AI) are being deployed in humanity’s hunt for the silver bullet to cure COVID-19. What may come as more of a surprise is that the solution may not be a shiny, new technology but instead an existing drug or molecule, repurposed to fight this novel disease. In other words, our silver bullet may just need some repolishing.
Repurposing a drug to treat COVID-19 can have a number of benefits, as argued by Professor Davenport, of Cambridge University’s Department of Medicine:
These [repurposed] treatments will have already been shown to be safe and so, if they can now be shown to be effective in COVID-19, they could be brought to clinical use relatively quickly''1.
One such drug is Remdesivir. Originally developed as a treatment for Ebola, it did not prove effective enough to reach the clinic. Now, however, it could be a powerful tool in treating COVID-19 and has been assessed in Phase 3 trials by Gilead Pharmaceuticals.
Whilst Remdesivir may seem a relatively straightforward repurposing – already being an anti-viral drug - a number of less obvious candidates are showing promise. For instance, pancreatitis drugs Nafamostat and Camostat are undergoing Phase 2/3 trials for patients with coronavirus. Even Fluvoxamine, an antidepressant used to treat obsessive compulsive disorder, is undergoing a Phase 2 trial. If successful, these medicines have key advantages in abetting the pandemic: they already have regulatory approval, they are known to be safe and we already know how to make and distribute them at scale.
One of the key developments in our effort to understand SARS-CoV-2 was the early sequencing and public release of its genome. With this development, the scientific world was able to access the genetic code of the virus and turn from a public health problem, to a bioinformatics puzzle. In an age of genomic data, when a pathogen can be rapidly sequenced and investigated, a number of investigators are turning to AI and machine learning (ML) models to uncover chinks in genetic armour. These same techniques can be applied to drug repurposing.
There are a number of UK biotech companies taking this approach. AI drug discovery powerhouse, Benevolent AI, used its platform to identify a drug candidate capable of stopping the progression of coronavirus infection. Within a month they had identified the arthritis drug Barcitinib, which is now entering Phase 3 trials in the US. Cambridge-based Healx are another company making use of the vast amount of data now available in healthcare to discover new treatments with a focus on combination therapies. By stripping out human bias, and employing the processing power of an AI-implemented system, they believe they can discover connections between drug candidates and diseases that clinicians may not have even considered.
It’s clear that employing AI-led systems to discover new purposes for existing drugs is an exciting and promising business strategy. But how best can IP, and patents in particular, support such a business model?
Upon developing an AI platform, a company’s first instinct might be to file a patent application for their new AI drug repurposing platform. However, this might not be the best strategy. As we have discussed in previous articles, protecting AI-implemented inventions is not straightforward at the European Patent Office (EPO). Furthermore, an AI or ML method is likely to be constantly learning and evolving. A patent is therefore likely to capture a snapshot in time of the invention, but might ultimately fail to protect the commercial embodiment. How, then, could such a company build a sound IP standing?
Rather than patenting the central AI platform, IP protection could be focused on the output: the repurposed drugs. Such drugs could be protected in Europe using what is known as a ‘second medical use’ claim. Even if the drug is already known for treating a first disease, the discovery that it will be effective in a new treatment area may be inventive in its own right. Such patent applications are, generally, much less complex to describe than AI-related inventions and are therefore likely to be cheaper to file. The application would need to be supported by some data beyond the AI’s initial prediction, such as some early data from animal models of the disease. The benefit, however, is that the core asset of the business – the AI model – remains secret: you need never disclose your golden goose, just its eggs.
The attractiveness of drug repurposing business models is clear. The great challenges in pharmaceutical development boil down to two imposing stats: the average time to bring a drug to market is between 13 – 15 years, and it will cost $2-3 billion in the process2. Repurposing circumvents the drug development valley of death and, on average, costs $300 million. A company that delegates the heavy lifting to an AI platform is likely to take this order-of-magnitude saving even further. By taking drugs that already have safety data and regulatory approval, there is a shorter and more attractive route to market for the new drug. The drug repurposing company might then take the drug forward or, avoiding the risk of drug development altogether, license out its invention to a better suited company.
Clearly, huge progress is being made to tackle the coronavirus pandemic. Vaccines are being developed in months rather than years, and our understanding of the virus grows each week. It may be, though, that the early heroes in healthcare are those scavenging on the scrapheap of medicines, rather than developing new ones.
Many thanks to Matt and Michael for sharing their thoughts on this fascinating topic. If you have any questions for them, please don’t hesitate to get in touch.