OUNTAIN VIEW, Calif. — Dr. Jessica Mega, a star cardiologist at Harvard Medical School, swapped coasts early this year to become medical director of Google Life Sciences — the Silicon Valley giant’s new foray into health and medicine.
Before taking a leave from Harvard, Mega led groundbreaking clinical trials that used genetic and molecular signals of disease to study customized treatments for heart patients. That put her in the sweet spot to lead the new firm’s ambitious quest to analyze genomic, molecular, and imaging big data from 10,000 volunteers to figure out what it means to be healthy — the so-called baseline study.
Google Life Sciences — spun off when Google recently reinvented itself as Alphabet, a holding company for Google itself and affiliated entities — has generated buzz in the medical community. Its experts try to turn blue-sky ideas into products by cross-pollinating medicine, engineering, and data science.
They’ve already produced a contact lens for diabetics that continuously monitors glucose in tears. Another prototype “paints” nanoparticles with substances that detect cancer or other abnormalities. As the particles flow through the body, a wrist band — think Fitbit’s smarter sibling — peers into blood vessels to read their findings.
Mega recently sat down with STAT at Google’s offices to talk about the baseline study. The conversation that follows has been edited and condensed.
You moved from the grand tradition of Harvard to the disruptive world of Silicon Valley. Have you felt any culture shock?
Boston has a rich tradition in life sciences, but it’s clear to me — and the reason I’m here — is that technology is moving quickly into health care, and Google Life Sciences is taking it really seriously.
I’m normally around physicians and patients. Now I spend my days with amazing engineers. The things you hear around here are “try to fail fast,” and “let’s just try ideas.” What I’ve taught myself to do is first say “yes” and try to be very open, then get analytical and move to a point where we’re being strategic and tactical.
Every day I come in assuming I’m going to learn something new, that I’m going to have to figure something out. For people who like that dynamic environment, it’s really fun.
How do you define “baseline”?
For example, one day a patient comes to see me, gets a full checkup and says, “Doc, thanks so much.” The next day they may, unfortunately, have a heart attack. It’s not that you’re healthy until the switch goes on, then you’re unhealthy. There’s this transition. There are good technologies out there now trying to understand what’s going on at the blood-vessel level — before someone actually comes in with a heart attack. To figure out which one of those early signals is most important will be very helpful.
We’re currently in a very active pilot phase to try to get all the pieces to come together (for that kind of challenge) — trying to combine deep molecular data, clinical data, imaging data and patient engagement. As you integrate it, you start to look for patterns. If we could figure out platforms to seamlessly integrate all of that information, then the contribution of baseline is beyond just our study. That’s something I think places like Google and Google Life Sciences will be particularly good at — using machine-learning algorithms and trying to come up with new observations.
We’ll try to understand the fundamental building blocks — the systems biology for what it means to be free of overt disease, what we call “healthy.” Then we’ll follow people to try to capture the early signs of that transition, with the goal of trying to prevent disease as we go forward.
Is it really possible to understand “baseline human health,” given the wide variations between people?
I think it’s more knowable than ever before. There are different tools available now, even compared to a few years ago. It sometimes takes time to understand the best way to use these tools.
But you bring up a really good point: Biology can be humbling, and people can change. That’s why even in our study we want to follow people over time. Are there intrinsic variations in individuals over time even if they don’t manifest certain evidence of disease? Until we start to look, we won’t know.
Sometimes the use of increasingly sensitive diagnostic tests or devices results in ambiguous choices and overtreatment. Prostate cancer comes to mind. How do such contradictions inform your approach?
That’s a concept that I think about a lot. What signals are actionable and what signals are noise? With certain signals, such as an irregular heartbeat, there’s consensus on what to do.
A great epidemiological example is, if I have matches in my pocket then I have a higher risk of having lung cancer. Is it actually carrying the matches around, the weight in my pocket, that leads to lung cancer? No! It’s because I used the matches to light a cigarette. So you come up with all these observations, and how do you know that by acting on them you’re going to change the course of disease? You have to tease out whether or not you are looking at something that’s causal.
Do you worry that such amazingly precise, data-intensive monitoring might needlessly drive up the cost of health care?
The way I think about it is trying to understand more about a given individual so they get the right treatment, get the right medications, and avoid the side effects. We’re trying to figure out ways to help empower people so that they don’t need to spend as much time in hospitals.
People don’t want a lot of unnecessary, expensive, cumbersome, inaccurate tests. But we’re working to come up with things that provide actionable information. People ask if this will be something that only a handful of people can use. The hope is that they will be scalable. We’re a bit early in that mission, but it’s something we take seriously.
How long have you committed to Google?
I’m here and I’m on the mission, and we’ll see where it takes me.