More than 40% of start-ups active in drug research use artificial intelligence to scour chemical repositories.
In ancient Greece, Hippocrates tried very early to introduce scientific rigor into the “art” of medicine. Many intellectual reflections over the thousands of years that followed made medicine the scientific field we know today.
The drug creation process has become a large field of investigation, characterized by very complex, time-consuming and expensive multidisciplinary methods carried out by a multitude of local, national and international organizations.
Drug discovery and design has come a long way from the days of iteratively applying known natural toxins (such as those found in fungi or plants) against specific diseases until a therapeutic effect is observed. Today, biology is digitizing at a dizzying pace, thanks to the recent drop in the cost of gene sequencing.
Using AI, a team of researchers at BenevolentAI identified baricitinib as a potential treatment for COVID-19 within four days.
However, the vast amounts of machine-readable data present both an opportunity to gain new knowledge and a formidable challenge, as it becomes increasingly difficult to exploit it with ever-increasing volumes of data. It is also more complex to maintain an overview of key developments in adjacent research areas, which can often be useful for drug design. Recent studies have highlighted this point, estimating that nearly 80% of medical data remains unstructured and unused after being created (Kong, 2019).
After enabling significant progress in other markets such as cloud computing and cybersecurity, can AI play a leading role in drug discovery?
Progress thanks to multidisciplinary open archives, HAL
In theory, drug discovery and design is precisely the kind of challenge that intelligent automation should lend itself to. For example, the number of possible permutations of drug molecules is around 1,060, which poses an interesting challenge for AI, which can be trained to recognize potential lead compounds and validate drug target and structure design. This task can be prospective and retrospective.
The power of AI is illustrated by the fact that, in just four days, a team of researchers at BenevolentAI identified baricitinib as a potential treatment for COVID-19. This Eli Lilly drug, typically used to treat rheumatoid arthritis, can attack both the COVID-19 virus and the body’s inflammatory reaction to it. It was the first time AI had discovered an existing drug to attack a new problem.
Having analyzed its tangible benefits, many companies are taking advantage of intelligent automation. In 2020, for example, Pfizer couldn’t afford to automatically filter any of its libraries containing data on 4.5 billion commercially available compounds. Today, it can scan the entire database in 48 hours, greatly accelerating its ability to identify potential new drugs.
Currently, AI is not expected to replace the human experience. Instead, AI is seen as a way to improve it.
According to Deep Knowledge Analytics 2019, there were more than 170 research and development (R&D) companies with AI technology around the world and 35 major R&D centers using AI. A 2019 Deloitte survey found that over 40% of drug discovery startups use AI to scour chemical repositories to find potential drugs, 28% use AI to find new drug targets, and 17% use it for computer-aided molecular design. . Miraz Rahman, professor of medicinal chemistry at King’s College London, estimates that within ten years, all major pharmaceutical companies will have incorporated AI into drug design.
Trust however check
It is important to note that AI is currently not expected to replace the human experience. Instead, AI is seen as a way to improve it. Subject matter experts are essential in defining the data for AI analysis and providing peer review and verification of the results. Also, like any powerful tool, AI can be used for harmful purposes if left unchecked. In a recent demonstration, an AI model was trained with a set of starter molecules and tasked with figuring out how to adapt them to make them increasingly toxic. The result was worrying: within hours, the model had proposed more than 40,000 potentially dangerous molecules.
Drug research is just one phase of the larger new drug approval process. As with discovery, other parts of the process have idiosyncratic inefficiencies that can be improved upon. Gene sequencing continues to improve in terms of speed, accuracy and cost. Illumina is the dominant player in the industry, but newcomers like Oxford Nanopore continue to drive innovation. Cell manipulation is seeing dramatic advances thanks to companies like Berkeley Lights. In addition, Genmab has made excellent progress in the area of antibody testing. Finally, clinical research organizations like Icon allow large pharmaceutical companies to outsource some of the most demanding tasks and focus on the most complex research. Each part of the value chain improves and contributes to a greater result.
Drug discovery remains an important part of the overall treatment development process and potentially lends itself to a greater role for trained automation. With the right balance between man and machine, along with the proper controls to ensure rigorous task execution, AI is expected to play an increasingly important role in how we discover new therapies. The future of new drug research is promising. Hippocrates would be proud of how far he’s come.