Reframing the conversation: Drug discovery and AI at JPMorgan Chase Healthcare Conference
As usual, there’s a lot of news coming out of the 38th Annual JPMorgan Chase Healthcare Conference. And machine learning’s ability to support the pharmaceutical research and development process is no exception. There have been dozens of presentations on this very topic, and even more companies pitching the promise of artificial intelligence (AI). But, one thing is very clear: More education is needed about how AI can support pharmaceutical research.
The JPMorgan Chase conference is centrally a finance and investor meeting. And what we’ve seen so far is that healthcare investors continue to bet on AI startups at an unprecedented rate. This trend is not slowing. However, from our vantage point, there are few companies that employ an approach that has high potential to find new treatments, and more importantly that considers the unique needs of biotech and pharma companies.
Investments correlate strongly to hype. Too many investors, startups and drug discovery companies don’t fully understand the role of machines to support the drug discovery process. But they know there is vast potential. This isn’t specific to pharmaceutical discovery. There is still a lot of hype around AI in general, from diagnostics to patient care.
I fear the lack of education can have unintended consequences. For one, when any company ventures into drug discovery without the necessary knowledge about AI’s benefits and limitations, it’s very hard to set expectations and even harder to deploy AI in a way that supports the varying and unique capabilities of different drug discovery companies. This could result in a move away from AI because of bad experiences in delivering value. We know from experience that when AI is used to support the drug discovery process in the right way, it is effective and produces potential targets and medicines that may make a difference to patients.
Every company and every research team are different. In fact, some teams require process improvements that may not be best served by AI. For them, it is simply not a strategic fit and there may be traditional means that are more effective in solving efficiency issues. We may be the only AI-driven drug discovery company that would not hesitate to recommend a way forward that does not involve AI.
For others, there’s a pain point in the process that AI could help to fix or make more efficient with data. The rest of their process may be working well. Fundamentally, AI is not changing the proven R&D approach but intervening at a point where improvements are needed. An example I often use is Alzheimer’s disease where billions of dollars and years of time have been wasted on drug targets that prove ineffective or inefficacious. AI, using data, has the potential to screen out targets that may not bear fruit.
And finally, there are some companies that need a systemic change. This often requires utilizing AI for many parts of the discovery process. The bigger point is that utilizing AI should be aligned with a company’s current needs and culture.
What few AI-driven companies will tell you is that AI is not a tool to replace humans or even a broad solution. We are far from a future where this is even possible. AI itself is specific to parts of the process, and using AI is specific to a company’s needs. Drug discovery is complex, and it requires critical thinking and experience. We have to reframe this conversation and look at AI as one valuable approach that can provide better data and ultimately better outcomes for patients, at the right time and with the right teams. Not as replacing the whole or any part of the discovery process.
The good news is that despite the hype, there is a clear sentiment at this year’s JPMorgan Chase conference that AI tools are a huge asset for drug discovery. What we need now is better education and knowledge on the potential of AI, and more importantly agreement that both the AI data side and the pharmaceutical discovery side need to work together as one team, not a in silo.