I’m actively working on advancing an innovative technology solution in biotech, AI for drug discovery. So, naturally, Matthew Herper’s article “Here’s Why We’re Not Prepared for the Next Wave of Innovation” for STAT caught my attention.
I quickly found myself agreeing with many of his points. The message I’m left with is that to harness all the innovation happening in biotech today, we need to make data more available in an organized fashion. We then need to be smarter and more creative in our ability to do more with that data, which includes using it to inform what further data we generate. While Herper focuses on clinical research, it’s clear to me that the same trends are also true in earlier stages of drug discovery.
In the article, renowned physician Peter Bach said, “We have constructed a system where cloistering information is the path to profits.” This quote really resonated with recent conversations we’ve had at Aria about how and why the current R&D system incentivizes proprietary data. The quote itself sounds innocuous until you remember that this “information” is either patient-derived or at the very least, relevant to understanding diseases. By cloistering it, companies, and to some extent academics, are inadvertently slowing down the advancement of new medications.
I’m not so naïve to think that all data should immediately be open access. There are important and thorny topics to work through regarding privacy, and we must absolutely be incentivized (i.e., “a path to profits”) to perform difficult research. However, I would argue that if biomedical data was generally more available, as researchers, we could focus our time on forming more targeted hypotheses and running the specific experiments (or clinical trials) we need to fill in the blanks. The idea of making privately generated data more available may sound utopic, but there is some growing momentum with big pharmaceutical companies sharing their data in precompetitive coops (e.g., Pistoia Alliance) and government generation of these sorts of data (e.g., UKBB). I believe this trend will not only continue but accelerate.
With more readily available biomedical data, I believe the emphasis for innovators and companies will shift from data generation to data analysis. It will become most important to do more with the data you have than to simply generate more data. Of course, I’m biased because I’m part of a team at Aria that is doing exactly that. We’ve built Symphony™, a proprietary AI platform to integrate and interrogate completely unconnected multimodal data. This unique ability means that in our discovery process we’ve shifted the importance from having special data, to having special analytical abilities. And by making our proprietary advantage our AI platform, we can use it again and again across diseases and without having to generate more data.
This takes us to another point that I really appreciated from Herper’s article: We shouldn’t just focus on using technology and innovation to further speed up (or I would say iteratively improve) the processes we already have. In Herper’s example, he points out that a lot of innovation has been spent on making clinical trials faster and more efficient at generating large amounts of data, but you could argue that the larger societal impact would come from making clinical trials easier to run.
Throwing new technology at existing processes to make them faster and more efficient isn’t inherently wrong by any means but doing so is unlikely to bring about substantial change. Take for example the work my team and I have done while building Symphony™. We’ve certainly spent time ensuring processes are fast and efficient, but what has been repeatedly far more impactful is to rethink the problem we’re trying to solve and then build a solution to address the problem head-on — even if that means throwing away previous work.
A mindset shift like this will be necessary for our industry to continue what Herper calls our golden age of drug discovery — so named because the last decade has seen a drastic increase in the number of medicines approved by the FDA compared to previous decades. Certainly, the last ten years have seen some incredible technological progress, but what this means is that the remaining unmet need will have even more complex problems and complex biology. To tackle these tough problems, we are going to need different ways of discovering drugs rather than just a faster version of what we’ve traditionally been doing.
I don’t think I can wrap up better than Herper himself did “[i]t’s not just that we need to do what we are doing better. …The world’s going to be transformed — we can’t let our thinking about it fall behind.”