In this founder talk, Helmy Eltoukhy frames cancer as a global health crisis—1 in 8 deaths worldwide, with incidence rising as populations age. He argues that progress has been hampered by a lack of molecular information across three stages: pre‑cancer (early detection), active cancer (rapid evolution and treatment resistance), and post‑cancer (survivors fearing recurrence). He traces four technological building blocks that made liquid biopsy possible: the 1958 integrated circuit (digital data processing), the 1982 discovery that cancer is a genomic disease, the 1994 finding that tumors shed mutated DNA into blood (ctDNA), and the 2001 completion of the human genome draft that drove sequencing costs down to ~$1,000. Using these advances, Guardant Health launched Guardant360 in 2014—a comprehensive blood‑based genomic test that replaces costly, risky tissue biopsies (average $14,000, 1‑2% mortality). He shares a patient story (Mark) whose stage‑4 lung cancer was found to have an actionable EGFR mutation via Guardant360, enabling targeted therapy and improved outcomes. By 2017 Guardant had processed ~40,000 samples; a machine‑learning back‑end continuously improves test performance, achieving a ten‑fold boost after 25,000 cases and projecting another order‑of‑magnitude gain. Leveraging economies of scale, Guardant launched Project LunaR (May 2016) to develop a sub‑$1,000 blood test for early detection and recurrence across five cancers (lung, ovarian, breast, colorectal, pancreatic). The key barrier to early detection is specificity—current screening tools (low‑dose CT, mammography) have false‑positive rates of 90‑96%. Guardant is amassing the world’s largest blood‑based cancer genomic database (~40,000 profiles across 40 cancer types) and applying sophisticated machine‑learning to reclassify cancers into micro‑subtypes, dramatically improving specificity. He projects that with 100‑200k samples, a breakthrough early‑detection test will be feasible, turning every Guardant360 patient into a data point that brings the goal of conquering cancer closer.
Chapters
Yeah, thank you, Joshua, for the kind introduction. It's great to be here today. So, 1971, 46 years ago, the US Congress declared war against cancer. Where are we today? Well, there are 26 new cases globally per minute, about 6 since I started speaking. 1 in 8 deaths worldwide, you know, caused by cancer. In the U.S., it's actually 1 in 2 men and 1 in 3 women. Obviously, it's one of the most important health challenges of our time today. And it's not just a disease of the old. And as the world ages and it gets more developed, it's actually getting worse, not better. And so why has progress been so slow? Let's take a step back and then really kind of zoom in on the problem here. If we look at the pre-cancer setting, early detection, we all know that outcomes improve the earlier we find the disease. For active cancer patients, Because cancer evolves, you know, so rapidly, it's hard to stay ahead of the disease, and cancer seems to evade whatever treatment we throw at it. And in the post-cancer setting, there are tens of millions of cancer survivors living in fear every single day that their cancer is going to come back, and they don't really have any test that can give them quantitative peace of mind. So what's the commonality here between these three vignettes? It's the fact that there's missing underlying molecular information across these three stages. Essentially, that missing molecular information is causing this slow iterative feedback loop we have in terms of making progress against the disease. So we're data-starved in terms of the underlying state in each of these different scenarios. So how do we fill that gap? Fortunately, the solution has been developed over the last few decades. First building block of that solution really started in 1958. With the invention of the integrated circuit and essentially the inexorable march towards the digital age which we now enjoy à la Moore's Law, where we can now process terabytes of information with relative ease. Second building block occurred in 1982 with the discovery that cancer is a disease of the genome. It's essentially mistakes in the genetic code of individual cells that largely give rise to the disease. The third building block happened in 1994. Essentially, this was a discovery by Maurice Strone that genetic sequence or those mutated DNA are not locked away inside tumors, they're actually shed by these tumors into the blood. If we could actually unlock those signals from a simple blood test, which is what is known as a liquid biopsy today, it would be pretty powerful. Unfortunately, the tools didn't exist in 1994 to unlock those tiny signals. And then in 2001, the final building block was essentially the completion of the draft human genome sequence, and the ensuing investment that happened over the 15 years to bring down the cost of sequencing a human genome by nearly a millionfold. You know, now it's a $1,000 genome. In a few years, it may even be a $100 genome. We'll see if that happens, but it's pretty exciting that we have that tool now. And so let's fast forward to 2014. What's depicted here in this picture is a lung biopsy. That was essentially the procedure that was used to essentially get important vital genetic information from cancer patients. So that they could be matched with the most appropriate therapies. In the US, the average lung biopsy costs $14,000, has a 1-2% mortality rate, and a 19% complication rate. So it's pretty grim outcomes for such a simple procedure, seemingly simple procedure. Later that year, we essentially developed a blood test, or the world's first comprehensive liquid biopsy, which could enable physicians to get access to that same genetic information, but through a couple teaspoons of blood. Needless to say, this test really revolutionized the cancer space. It was one of the fastest adopted tests in oncology, and now thousands of oncologists order this for their patients every single year, largely because of patient stories like Mark's that you're about to see. If we can cue the video, that would be great.
I just have a sense of freedom flying through the woods, up and down hills. It's, it's a big kick for me. My name is Mark. I'm a professional photographer in Atlanta, Georgia. I've got 3 daughters, very happily married. I'm an avid cyclist, and I've got stage 4 lung cancer. My scan indicated a 5.1 centimeter in length tumor and metastasis throughout my chest cavity and a cancerous lymph node on my liver. Once it started to sink in, I just knew that it was, it was going to be a tough fight. And I was, I'm prepared. Genetic mutations are what treating lung cancer is all about with successful outcomes. I was told that there was nothing there that's targetable. That was probably the biggest blow. Blood biopsies, or liquid biopsies as they call them, are kind of on the horizon. I said, I'll be happy to give up 2 vials of blood if we might be able to find something. And sure enough, we got the good news that the Guardant360 test had revealed an EGFR mutation. I about dropped the telephone. I'm enjoying dramatic results. With this cutting-edge testing, we can cut out a lot of the guesswork and match mutations to known effective therapy. I had a lot of things in my life to look forward to, one being my daughter's wedding, and I'll have a high school graduation coming up for my youngest daughter. I've got all the tools that I need to fight this, and I am grateful every day.
So, um, yeah, and Mark is still doing well. We ran another test on him. His cancer was coming back, and he was matched with another therapy that is now putting it at bay for the time being. And so it's really exciting to see in just 3 short years going from essentially driving blind, flying blind, to be able to proactively and adaptively manage the disease in late-stage patients. So where are we going? It's 2017 now. We've processed nearly 40,000 patient samples, and it turns out because we have a machine learning back engine in terms of, you know, how we process the data. Our test actually gets better with every sample we run, with every patient we help, the performance improves. And so now that when we hit 25,000 patients, we actually saw a 10x improvement in terms of performance. And now we're at the point where we're projecting in the next 18 months we'll get to even another, maybe, order of magnitude improvement in terms of where the technology will be. And what that will enable is not just use of this technology in active cancer patients, where the signals are larger, but using it for the other two settings I mentioned, for cancer survivors, as well as essentially for every single one of us for early detection. So let's, let's focus on early detection a little bit. So, you know, we launched a major program in May of 2016 we call Project LunaR. Essentially, because of the economies of scale we've built with the technology, we're essentially now able to develop a test using the same platform that would be a low-cost test, be priced under $1,000 for early detection, recurrence detection, initially focusing on 5 cancer types: lung, ovarian, breast, colorectal, and pancreatic cancers. And we're fortunate to have partnered with over a dozen medical centers across the world some of them listed here. And early detection is not a new challenge. People have been trying to solve early detection of cancer for many years. And so what's the challenge there? The challenge is one of specificity. That's really been the Achilles heel of tests that have been developed for that indication. When you look at, in the US, high-risk screening for lung cancer, is now done largely by low-dose CT scans. Low-dose CT scans, even in that high-risk population, have a 96% false positive rate. That is, 96% of the time they see a nodule that is benign, is not cancer. Mammography has over a 90% false positive rate. And so specificity is really the problem. It's almost like, you know, we're trying to detect something that we don't understand. We don't even know what it truly is. Just to put this in, you know, analogy that may make more sense or may resonate more, this is as if, you know, physicians were trying to fight infections without knowing what the associated virus, you know, is or were— was. And, you know, that's just crazy to think that we haven't really defined cancer. Cancer is really an agglomeration of tens of thousands of different, you know, diseases. And, you know, to fast forward to 2017, how do we solve that specificity issue here? And so we've been quietly amassing the world's largest blood-based genomic database in cancer. We now have almost 40,000 genomic profiles of cancer patients across 40 different cancer types. And what we're doing with that data is we're applying sophisticated machine learning algorithms to this to essentially reclassify the disease, redefine the disease in unprecedented resolution down to very, what we call, microclasses of subtypes of cancer. And what we find is, as we accrue more data, that our specificity improves dramatically. And we project when we get to, you know, 100,000, couple hundred thousand samples, will be at a specificity that is sufficient for, um, really a breakthrough blood test for early detection. And so in this way, every Garden 360 active cancer patient we help in turn helps bring us one step closer to a breakthrough test for early detection, and at the same time helps bring us closer to conquering cancer with data across all stages of the disease. Thank you.
