The Quantified Computer Scientist: Larry Smarr on the Future of Medicine
James Temple, Re/code
Welcome to De/code: A new series of conversations with leading thinkers in health and science.
Computer scientist Larry Smarr is standing in front of the “Big Wall.”
The tiled display of LCDs spans the length of a room in Atkinson Hall, a modern glass-and-steel structure on the University of California, San Diego’s palm-covered campus.
Cast across the screens are Smarr’s insides, or at least the closest you can get when breaking down the microbiology of a human body into data. There are some 150 variables in boxes color coded against their divergence from the norm: There’s magnesium in green, lysosome in yellow and lactoferrin in red.
That one’s trouble.
The protein is a critical player in the immune system, and the spike rendered in stop-light red, signaling levels more than 100 times greater than normal, is a strong indicator of Crohn’s disease or ulcerative colitis.
Smarr spotted the issue years before any symptoms developed and well before any doctor diagnosed him, because he has meticulously tracked his personal health information for years.
It began when he relocated from Illinois to San Diego, transitioning from a professor of astronomy and physics at the University of Illinois at Urbana-Champaign to a professor of computer science at UCSD. Southern California’s hard bodies inspired him to shed a few of his Midwest pounds, and he began to closely track his food intake and exercise.
But he soon went beyond the average calorie counter. In the last five years, he has submitted dozens of blood and stool samples. He wore a headband that monitored his sleep patterns for nearly two years. He has undergone MRI scans and analyzed his DNA. He tracks his daily steps with a pedometer. And he amasses and studies all that data over time, watching closely for shifting variables.
Of course, Smarr is better positioned than most to analyze, interpret and make use of that data. He shifted from astrophysics to computer science after recognizing the promise of supercomputing for unraveling the mysteries of the sprawling universe. Now he’s applying similar tools — turbocharged by a few decades of Moore’s law — to the microscopic world within our bodies.
Smarr is the founding director of the California Institute for Telecommunications and Information Technology, a partnership between UC San Diego and Irvine.
He’s leveraging the considerable computing resources at the San Diego Supercomputer Center to study the potential for applying more and better data to personalized medicine, in particular by researching the unique makeup of organisms within our bodies known as the microbiome.
(This is all when he’s not working on developing the next generation of cyberinfrastructure.)
Among other things, Smarr is trying to understand what has gone wrong in his own body and how best to fix it, exploring the potential for cocktails of probiotics to replace critical missing bacteria.
“Basically, we will have personalized doctors with us at all times, instead of two 15-minute visits a year.” — Larry Smarr
While the information revolution has transformed retail, media and even medical research, Smarr believes it has largely skipped over the practice of health care itself.
But he believes we’re on the cusp of a radical transformation as data, software and sensors replace frenzied reactions to acute symptoms with ongoing monitoring and maintenance of health.
Last month, standing on the dais in Atkinson Hall, Smarr discussed how we might get there and what it will mean, in what amounted to a half presentation, half interview with Re/code. (The discussion has been edited for length and clarity.)
Re/code: You were quoted saying that these new tools will mean we have “for the first time in history, a scientific basis for medicine.” What did you mean by that?
Larry Smarr: Doctors are incredibly skilled at looking at the epiphenomena of our bodies, the symptoms, and deducing what must be going on inside you — instead of what we’re talking about, which is just directly measuring biological variables inside you.
I think you’ll see more disease stratification, which you’re already seeing in cancer. Before we could genetically sequence the solid tumor, you’d say, “Let’s try this chemo, oops that didn’t work, let’s trying this chemo.” Which is a horrible thing to have to go through.
Read other stories in the De/code series:
- D-Wave CEO: Our Next Quantum Processor Will Make Computer Science History (Video)
- Artificial Intelligence Raises New Hope for Cancer Patients
- Medicine’s Big Problem with Big Data: Information Hoarding
- Self-Assembly Required: One Scientist’s Bid to Build Cancer-Killing Nanorobots
- Beyond Evolution: Scientist Designs Life From Scratch to Combat Disease
- Elizabeth Kolbert on How Tech Can — And Can’t — Tackle Climate Change and Extinction
- The Quantified Computer Scientist: Larry Smarr on the Future of Medicine
- One Scientist’s Bid to Debug Human Software
But now what they do is sequence your tumor cells, in which your DNA has some specific mutation. So by knowing what mutation that is, you know which chemotherapy will have the best chance of being effective. That’s already sweeping the cancer world.
I’ve come to love bugs and their relative characteristics. The microbiome … is 90 percent of the cells in your body. But the remaining 10 percent that is human is what medicine is about today.
When I had my first cosmology course, in 1968, we thought the universe was made up of stars and galaxies and stuff you could see. Then it turned out there was dark matter and dark energy, and we’ve found all the visible matter is a few percent of the entire mass of the universe.
We’re about to go through a similar revolution in medicine, where we find out the microbiota world is the dark matter of the body. It’s been completely ignored up until now — and it’s going to have huge implications.
Are there technological hurdles that remain to getting to this end vision? Improved data processing or a better understanding of what the data means?
I’m pretty impressed with things like Fitbit and Nike FuelBand and Jawbones and so forth. Typically, when someone does get one of those things they just see how little they move and they greatly increase their amount of steps per day, which is one of the best things you can do for your health.
This idea of personalized biofeedback is one of the fastest growing areas. But if you’ve ever had the experience of taking your numbers in to see your doctor, whether it’s your genome or Fitbit numbers, they say, “well, there’s nothing we can do with this.”
So where do you get advice other than whatever random stuff is on Google?
Well, this is where [IBM’s artificial intelligence system] Watson comes in. The whole rise of artificial intelligence and what’s now called deep learning is going to be critical.
Imagine you’ve got Watson. In addition to knowing everything that’s in the medical literature and all the text books and everything else, it has your numbers that are being recorded all day long. How much you’re moving. Pretty soon your continuous heart rate. Your food. It learns you as well as Google learns what advertisements to post for you, which continually refines its view of who you are.
That technology is at the heart of the medical revolution. Basically, we will have personalized doctors with us at all times, instead of two 15-minute visits a year.
On the issue of the Fitbit and weight loss apps, I’ve talked to people working in this area and what you’ll hear is there’s a magic number — three months — where people stop using them.
My take, having done all this stuff for over 10 years, is the main reason people do that is not that they don’t think it’s useful. It’s that they’ve learned what there is to learn about their personal lifestyle and then they’ve embedded that as changed behavior. So now they walk everywhere to do their shopping. They park as far away as possible. They don’t take the elevator anymore.
Roughly you’ve learned what to take out of your diet and what to put into it. And that’s a huge win.
I guess my question was just whether or not there are some advances to be made still. Say health sensors that can passively track our calorie intake as opposed to making you enter them in.
The killer app is going to be, and I don’t know whether it will be technically possible, if you could get to a continuous glucose monitor that didn’t require you to puncture your skin into a blood vessel. That would be revolutionary. (See Google X’s effort to create contact lenses that might do just this.)
The fundamental thing that is driving the obesity epidemic is spiking your glucose constantly and bringing in insulin.
So for instance, the average American drinks between 500 and 600 12-ounce sweetened drinks a year. I have none. So that means they’ve spiked their glucose 500 times a year more than I have.
Well, your pancreas has so many shots of insulin per lifetime — that’s not quite true, but essentially. So if you’re using it up gratuitously, before long you end up insulin resistant and then all kinds of bad things happen to you.
The [Centers for Disease Control and Prevention] estimate that a large fraction of chronic diseases are preventable by different lifestyle choices. The number one cause of preventable chronic disease is smoking. Yet there’s still 18 percent of the U.S. population that smokes, according to the latest numbers.
Two-thirds of the country is overweight or obese. Most of that is lifestyle choices because, in 1970, it wasn’t that way.
That brings up a question for me. Human nature being what human nature is, is more data going to be the answer when we can’t get people to listen to the data we already have?
I’ve given a lecture titled “Science is Not Enough.”
The hardest thing to do is behavior modification. The good news is our country is extraordinarily skilled at it. That’s why marketing of sugar — toxic sugar — and smoking works.
So here’s a number. In 1970, the average American consumed one pound a year of high-fructose corn syrup. Today, it’s more than 50 pounds per year per person.
Let’s look at our national brands, look at what pension funds invest in. We’re very good at ruining our health through marketing. So it’s not so hard to imagine that we could turn it around, like with antismoking campaigns, to use marketing to try to keep us healthy.
It’s not simple, but it is doable. I think the other thing we’ve got to do is get incentives in place. If I want to get automobile insurance and I do drivers ed, I get a discount. If I have high grades I get a discount. Even in things like life insurance, if I don’t smoke, I get a discount. Well, can those incentives be applied to lifestyle changes that could, say, turn obesity around?
You’re obviously an outlier at this point, in terms of your willingness to go to these lengths, to take this many tests and sort through the data. How will this become more common?
Let me give you a little bit of a metaphor. When I was founding director of the National Center for Supercomputing Applications, back in 1988, I bought a Cray-2 supercomputer. It was about $15 million. There were motor generator units that looked like something out of a 19th-century steamship engine room.
Well, anybody’s smartphone today has more computing power and more memory than that Cray-2, by quite a lot, and it’s basically close to free. So the whole point of getting to digital medicine is to get medicine on Moore’s law.
Genome sequencing is very much that. The first human genome cost $3 billion to $4 billion to decode. Illumina just announced a few weeks ago the $1,000 human genome. That’s a million-fold reduction. That’s two decades of Moore’s law in 10 years, maybe 15.
We’re not going to do [full human and microbiome genome sequencing] for everybody, but if we find some of these major relationships, we can make an inexpensive targeted test.
Here at UCSD, we’re setting up a clinical trial of about 150 people: 50 with Crohn’s, 50 with ulcerative colitis and 50 healthy. Then you’ll begin to see, in detail, which biomarkers (measurable substances in an organism) and microbiome ecologies are really indicative of which phenotype (the observable characteristics of a human) condition.
And it’s not just us, there are people all over the world doing this. By the time you’ve got a robust test, the technology cost has gotten so low that it can afford to be mass market.
So you’ll be able to take a Q-tip, put it across toilet paper after you use it, put it in a vial, put in the mail and you get this detailed breakdown of what’s there for, I don’t know, $89.
Or urine strips, little pH strips. If it’s black you know you’re acidic and if red, base-y. So you can monitor your health in a way that says, “I’m off,” or “I’m okay,” just like standing on a scale. It gives you some feedback.
That’s how I think we’re going to get there.
Health care today is about, you get sick and then you get health care. Instead, we’re going to keep you healthy so you don’t need health care.