Five ways AI is reshaping the biomedical industry

Around 90 per cent of the world’s data was generated in the past two years alone, presenting the biomedical sector with a significant challenge, said Professor Jackie Hunter


When it comes to the pharmaceutical industry—and drug development more specifically—the chances of success are slim-to-none. Many scientists and researchers work their whole life in this field, never successfully taking a drug to market, said Professor Jackie Hunter, CEO of BelevolentBio.

Professor Hunter herself has been lucky—throughout the course of her career to date she’s helped develop and successfully take a drug to market on more than one occasion. And now, as she heads up the biomedical arm of UK-based BenevolentAI, Professor Hunter hopes to make an increasing number of breakthroughs with the help of artificial intelligence.

“I joined Benevolent because I knew the industry was broken and things needed to be done differently,” she said in an exclusive interview at Web Summit in November 2017. “I know this is the right way to do it. I want to leave this industry with a lot of drugs that have made it into people successfully and making a difference to lives—that’s why I’m doing it.”

When we sat down with Professor Hunter, we spoke about the untapped potential of AI and how the technology is helping to close the industry’s information-innovation gap

1. Working smarter

Due to the limited capacity of a human brain to analyze information, there’s a finite amount of data that one person can access and process, Professor Hunter said. “This is why we need artificial intelligence—to help bring the most salient pieces of data to the attention of the scientist, because there’s no way the scientist could find it.” 

Beyond that, AI  is also capable of making connections that surpass human biases, she pointed out. “For example, our system found one paper relating a disease to a particular target.  I’ve worked on that target—I would never have related it to that disease in a million years.”

“Likewise, if you type that disease and that target into Google it will find the one paper, but if you type in what targets are related to this particular disease it would never find it because there’s only one paper.

“This is why intelligent systems can really help the scientists focus their attention on the things that are most relevant for them.”

2. Building a scientist-AI relationship

AI is not going to take scientists’ jobs, certainly not for decades yet, because it still needs the intelligence of the scientist to be able to add their layers of experience on top of the array of hypotheses and facts, said Professor Hunter. “For example, there’s a huge bottleneck in pathology. Artificial intelligence can allow the pathologist to look at things that are difficult, unusual, and really explore them thoroughly rather than having to trawl through lots of normal specimens.”

Read more: Robotics and automation in the medical sector

“It’s going to help make scientists smarter because they’ll be able to base their decisions on a much broader piece of information,” she said. In addition, it will allow healthcare systems to pay more attention to patients because resources will be focused on those patients and those areas that really need it, Professor Hunter added.

3. Closing the information-innovation gap

“Looking at cancer, you now have information about the patient’s genotype, you’ve got  information about how those genes are expressed—so you’ve got expression profiles, you’ve got information about what treatments they respond to, what mutations their tumor has, a whole range of huge information,” Professor Hunter pointed out. “Presenting this to an artificial intelligence system that can look for patterns that no human brain could ever find will help create new important discoveries about relevant medicines or new important mechanisms in a particular cancer.”

“There are a huge number of repositories—from tumor biopsies to patient data—it’s impossible to imagine how to pull this information together and analyze it without using artificial intelligence.”

4. Overcoming roadblocks

Another challenge is that it can be terribly hard to disambiguate a lot of the jargon, acronyms and synonyms within the biomedical corpus, said Professor Hunter. For instance, recognizing that ALS is motor neuron disease and not acid-labile substance, or that AD in this paper is atopic dermatitis and not Alzheimer’s disease, is a real challenge, she said. “We’ve overcome these challenges but it’s taken a lot of effort to do it.” 

“There are a lot of technical challenges unique to biomedicine or chemistry. That’s why we’ve spent quite a lot of time getting things right,” she added. “That’s our secret source, our know-how. We’ve built those very domain-specific capabilities up from scratch.”

5. Making a real difference

“BelevolentBio will have more hypotheses validated,” predicts Professor Hunter, pointing out that the organization has already demonstrated with one internal drug program, and are on track for another couple. “The first of our in-house programs is going to go into clinical studies next year and we plan to ramp up our pipeline,” she adds.

Read more: The future of drug discovery 

“We’re already talking to a number of pharma companies about assets they might have available for repositioning. From this, we’ll be able see whether we can come up with new indications.  I want us to grow and be a thriving mini pharma company—it’s biotech with a twist!”