Artificial intelligence has long been heralded as a game changer to the way we live and work. But is it the panacea that it appears to be?
“Long and laborious.” That’s how Peter Richardson, Vice President of Pharmacology at BenevolentAI, describes the traditional drug discovery process.
Using conventional methods, it usually takes eight years before researchers discover a chemical compound that is promising enough to begin clinical trials, and hopefully, form a new drug. “That’s because for every 10,000 compounds tested, only 10 to 20 make it into pre-clinical research and trials,” says Richardson, who is also the co-founder of the London-based biotechnology firm.
Yet when the COVID-19 pandemic struck early this year, BenevolentAI’s artificial intelligence (AI) platform identified baricitinib, a commercially available drug for treating arthritis, as a possible COVID-19 treatment over the course of a weekend. Baricitinib is now being tested in over 12 clinical trials worldwide, and has demonstrated improved outcomes in early clinical use on over 800 hospitalised COVID-19 patients.
How did AI play a role in this lightning-speed discovery?
Speed is of the essence in the search for a treatment against COVID-19. AI has proven to be a vital ally in this endeavour.
Richardson professes that the discovery was new territory for the company, which had never applied its AI platform to examine infectious diseases. “We believed we had a duty to mobilise our technology for the public good in the current global health emergency,” he says.
Identifying the drug, however, was anything but fortuitous. “We would never have discovered this so quickly without years of investment in building our biomedical knowledge graph,” explains Anne Phelan, BenevolentAI’s Chief Scientific Officer.
A knowledge graph is a database which graphically represents relationships between data points. BenevolentAI’s knowledge graph gathered data to uncover insights into the relationship between drugs and its effects on patients.
“AI’s power is unleashed when there is a large volume of raw data for the machine learning algorithms to feed on,” says Michael Zeller, Head of AI Strategy & Solutions at Temasek — a specialised team within Temasek’s Investment Group exploring how AI and data science can deliver better business outcomes. He notes that “companies that possess large sets of high-quality data across multiple domains have the potential to leverage AI as a competitive differentiator.”
Richardson believes that the speed at which baricitinib emerged as a possible COVID-19 treatment demonstrates the potential AI has in uncovering new treatments, especially for complex diseases where treatments either do not exist, or frequently trigger adverse side effects.
“Right now, drugs at best only work in half the patient population they are prescribed for,” he stresses. “Leveraging AI helps us better understand patients through more intricate metrics like medical history and genomics, crucially improving drug response rates.”
This precision medicine-guided approach paves the way for better diagnosis and optimised treatments. “It’s where I see AI having one of the biggest impacts on human health,” says Richardson.
"Companies that possess large sets of high quality data across multiple domains have the potential to leverage AI as a competitive differentiator." -Michael Zeller, Head of AI & Strategy Solutions, Temasek
Read the full story on Temasek Review 2020