Earlier this month, STAT national correspondent, Casey Ross, and a team of industry leading experts—including, Niven Narain, president and CEO of Berg; Christoph Lengauer, president and co-founder of Celsius Therapeutics; Mariana Nacht, CSO of Vivid Biosciences, and Donald Bergstrom, head of R&D at Relay Therapeutics—came together in Boston to discuss AI's role in drug development. I had the great pleasure of introducing the panel to a crowd of more than 130 attendees.
At CAS, we strive to help organizations leverage the right content and technology, in combination with human expertise, to access mission-critical scientific data that drives discovery and informs strategic investment decisions. So, I was eager to hear the panelists' perspectives on both the positive and challenging aspects of expanding technology into the drug development process. As I listened to the conversation, I found myself nodding in agreement. Many of the points they discussed are challenges we at CAS work to solve every day.
If you weren't able to join us in Boston, keep reading for key insights from the thought-provoking discussion.
Defining and applying AI
Ask three people what AI is, and you'll likely get three different answers. This point was reinforced by Nacht, who noted that everyone seems to interpret AI in their own way, and ultimately, how we apply it tends to define what it means to us. To her, it means feeding large amounts of data through an iterative algorithm that must continually be refined and redefined based on outcomes.
Narain shared that he defines it as when a machine makes an independent analysis without any human interaction. If you consider the core of the definition, he commented, you could argue that Microsoft Excel represents machine learning—you put data in, Excel processes it and you get a result. But as it pertains to healthcare, the definition is far more complex because verifiable results are crucial.
"It's really important [in healthcare] to not stop at the model," he said. "Because that begs the question—is the result just a correlation? For disease and health, [that's not enough], we need to get at causality."
At CAS, we believe the convergence of human expertise, technology and data lead to the best outcomes, so this point rang true to me because it aligns with the idea that AI alone isn't the answer to driving innovation. Human expertise is still critical to ensuring the right data structures and algorithms are in place and they must be continually fine-tuned to drive optimal accuracy and results.
Focusing on better data
A critical way AI helps researchers is by making sense of data and uncovering insights that would otherwise go undiscovered. For example, AI can crunch the numbers of huge data sets—everything from tissue samples and genetic mutations to the progression of symptoms. This helps researchers more clearly define their patient population, allowing them to make better decisions about how to evolve the development of a drug or administer it once on the market. But getting those insights requires the right data, as pointed out by Lengauer. He shared the example of a drug that delivered promising results for 40 percent of patients who tried it. "We have no idea why it didn't work for the other 60 percent of patients and why the 40 percent did respond. We can't stratify patients and we don't know at which point is best to give the drug—at diagnosis, as a last resort, during flare-ups? In drug discovery, you need to make drugs that work."
Lots of questions and little insight are frustrating realities for many researchers because in the world of drug discovery, obtaining a large amount of reliable data can be incredibly difficult. This emphasizes the need for standardized and industrialized data working in tandem with the latest technology. For example, in cancer clinical trials, there is a medium data problem. "It's too big for a human to see a pattern, but not big enough for algorithms to make sense of it," Bergstrom noted.
This brought to mind the work we do at CAS. Every day, we help clients access, organize and better understand their data assets by building more effective data management solutions. And the outcomes they're able to achieve confirm that quality, well-organized data is the key to driving innovation.
To hear all of the experts' insights, you can watch the full panel discussion on-demand. And if you want to maximize your organization's data assets to accelerate business, learn how CAS can help.