Elsevier and Every Cure: treating rare diseases with generic drugs — at scale

Elsevier is providing data and its algorithms to the project. Initially, Every Cure provided a list of 10 incurable diseases — ranging from iMCD to sepsis to autism. Elsevier’s Professional Services Group customized algorithms to follow a disease path for what drugs could be repurposed for what diseases. And here’s a fun fact: the algorithms tagged the same drug in seconds that David struggled to find manually to treat his iMCD. 

We decided it was an excellent time for a chat with David about his compelling mission. 

How did you react when Elsevier’s algorithm found in seconds the drug against Castleman Disease that saved you? 

I was amazed. It was really exciting because our goal is to make these connections in the most systematic and fastest way possible so we can treat as many patients as possible. My own discovery took a lot of time and effort, and things really needed to line up for me. So obviously this was super exciting. 

There is, however, a risk of bias in these results. Many years have passed since that discovery. More new information is now available in the literature that could lead one to believe that that drug could be useful. So I’d be curious to go back in time and see if the algorithm would have given it as high a score back then. It’s hard to know. But of course I was still amazed.

It’s also an exciting time in terms of funding —

Yes! I was able to just announce at the Clinton Global Initiative 2023 Meeting(opens in new tab/window) that we now have four foundational funding partners: the Chan Zuckerberg Initiative(opens in new tab/window), Flagship Pioneering(opens in new tab/window), Lyda Hill Philanthropies(opens in new tab/window) and the Elevate Prize Foundation(opens in new tab/window).

These are major players. What’s the core message you think attracted them?

That we want to identify more uses for existing drugs. That we want to cure patients with drugs that are already sitting on the pharmacy shelf. And we’re doing this because we believe that no one should suffer if there is a treatment out there. We also believe the medical community should work together to unlock these hidden cures.

When you have the opportunity to go deeper into your work, are there particular moments when people’s eyes light up as they grasp the fuller implications — and that these rather epic goals are actually doable?

Of course, Viagra always gets people’s attention. Most know it’s a repurposed drug — that it wasn't initially intended for erectile dysfunction but for treating hypertension. But fewer people know that it’s been repurposed further. Viagra is now also a lifesaving treatment for rare pediatric lung disease(opens in new tab/window). As a result, all these kids who were not making it to their teenage years can now live full lives thanks to Viagra. 

Or look at thalidomide(opens in new tab/window). It was first approved as a morning sickness treatment and caused thousands of birth defects in the late 1950s. But later, it turned out to be an incredible lifesaving drug for not only a common cancer, multiple myeloma, but also for leprosy. And leprosy and cancer couldn't be more different. But it turns out, the way thalidomide works, it can work on both diseases. 

And when you start sharing these sorts of insights and highlight how 80% of all FDA-approved drugs are now off-patent(opens in new tab/window) — and can therefore be generally made for much cheaper — people really start seeing the potential. 

How did the initial connection with Elsevier happen?

Naturally, I was very familiar with the important work that Elsevier does in bringing together data through journal articles and databases. And since we need amazing data sets, it seemed obvious to reach out. Dr Frederik van den Broek(opens in new tab/window) [Elsevier’s Senior Director of Professional Services] wrote back immediately to say, ‘We love what you’re doing, let's chat further.’ And when it turned out they were already working to apply AI to identify repurposing opportunities, it became an even more specific and exciting collaboration opportunity. 

What’s the current state of Every Cure? You have initial funding — now what?

Of course, this doesn't fully fund us by any means, but it allows us to hit go and work to get the right people in place. We just posted 10 roles, and I'm in the process of interviewing for Chief Technology Officer, Chief Medical Officer, and a number of other roles that are going to be really important for our success.

Over the last 18 months, we worked with world experts in artificial intelligence and data science to create an optimal data science strategy. And that feels rock solid, especially since none of this would have been possible five years ago — AI has just advanced so much so fast. 

And now thanks to our funding partners, we have the money to launch this strategy. And for this we need more data. And we’ve got groups like Elsevier contributing data towards our work, which is awesome. And we’re working to inspire others to donate their data to what we’re doing — since the more strong data we have, the stronger the predictions will be. 

Moving forward, what do you see as the biggest challenge? Funding? After all, you may be saving a lot of money and time in terms of R&D, and clinical trials are not cheap.

Yes, I’d say fundraising. This is because we are tackling the systemic problem of drugs not being fully utilized. And we can’t predict what drugs will be effective for what diseases. Traditionally, donors and philanthropists wanted you to save lives of a particular disease — pancreatic cancer or COPD or that disease that affected their child. And we may very well uncover repurposed drugs for those diseases. But it may very well be that there's another repurposed drug for another disease that might have a greater impact — and so we're going to go after that first. 

Our current funders understand this: they also just want to save lives — period. And we will need to find more groups like that. At the same time, we believe our strategy has a much greater chance at success than the traditional approaches of one-drug-at-a-time or one-disease-at-a-time. But it does mean that getting funds is going to be more challenging.

Elsevier’s focus has generally been more around supporting Pharma companies in their R&D. Here, we are working more directly with doctors. This is very exciting because in a way it’s a step closer to the bedside and being able to have a more direct impact. Do you have any concrete advice for collaborating with doctors instead of R&D teams?

I am very grateful and thrilled that Elsevier is willing to join us in thinking about things in a different way. And I’d say: keep doing that! Personally, for me and our team it’s been very liberating to think about how we can help people with the tools already at hand — and doing that without any filters around profits or focusing on a specific disease or coming up with a new drug. What happens when all that data from all those companies and organizations out there can be gathered simply to help people? It’s a whole other territory. 

And how about more specific feedback in terms of how the Elsevier algorithm can be improved? I guess there’s that universal AI-related challenge in terms of assessing how accurate a particular algorithm or prediction is.

Absolutely. And that's really difficult because the only way to truly assess if a prediction is accurate, we have to do a clinical trial. And of course, that’s not feasible, given that clinical trials cost millions of dollars, and we potentially have thousands of drug-disease combinations to assess. 

We all need to continue to work in finding new ways to, as they say, “validate the algorithm.” One way is to look at previously generated scores and see how accurate they line up with the clinical trials that have actually been done. Another way is to integrate as many datasets as possible into these models so they become as accurate as possible. We need to continue to work on that.

So while I think the Elsevier algorithm is very exciting, we will continue the process of refining it and assessing its accuracy. 

So you have this initial list of 10 drug-disease pairings. Any you are particularly excited about? 

In my opinion, this is really just a starting place — to test the algorithm and see how it compares to other algorithms. What I’m most excited about with Every Cure is a scenario where we don’t pick the 10 diseases upfront but the algorithm does. In other words, the data will tell us, among all drugs and disease combinations, what the 10 are that look most promising.

So is that what scaling looks like for Every Cure — that moment when the algorithm is actually doing all the heavy lifting? And what would be the timeline for that?

Yes, I think that’s what scaling looks like. And I'd like for us to have the top five or 10 drug-disease pairs within the next year. And at that point, we’ll also already be out there working with nonprofit organizations, government agencies and disease groups to try to get those clinical trials funded. So that way, we can really evaluate whether those drugs really do work in those disease areas. In other words, we will have our hits and we will be chasing them. 

And if I can jump ahead to three years, I want to have changed clinical practice for multiple diseases, meaning that there’s a new drug — or actually an old drug — being used in a new way for a disease that had a significant morbidity and mortality. In other words, patients will be doing better thanks to a drug that would otherwise have been just sitting on the pharmacy shelf.

What’s the single most important message everyone should know?

That the system has been working the way it was built to work: drugs being developed based on the areas incentivized financially. But because I survived Castleman thanks to one of these old generic drugs that everyone had stopped working on, it uncovered this incredible opportunity hiding in plain sight for all these years.

For me, it’s a dream of a lifetime to be able to partner with organizations like Elsevier, the Chan Zuckerberg Initiative(opens in new tab/window) and others. But we can’t do it alone. We need people, companies and organizations to join us. The horizon is bright, but there’s still a lot to do.