Q&A with Anthony Philippakis, Chief Data Officer of the Broad Institute
Anthony Philippakis, MD, PhD, is the Chief Data Officer of the Broad Institute of MIT and Harvard, where he leads the Data Sciences Platform team. He is also a cardiologist at Brigham and Women’s Hospital, where his primary focus is caring for patients with rare, genetic, cardiovascular diseases. In addition to his roles at the Broad and Brigham and Women’s, he is a Venture Partner at Google Ventures, focusing on life sciences and technology.
Anthony is a mentor at Insight’s Health Data program, offered in Boston and Silicon Valley, and every session shares his unique perspectives with Fellows on the intersection of tech and the life sciences.
How did you get to where you are now?
At the moment, I wear three hats in life: I’m a cardiologist, a venture capitalist, and the Chief Data Officer of the Broad Institute.
I’ve definitely had a bit of a winding road to today! I grew up in Arizona and went to college at Yale, where I was a math major. I loved pure math back then, and it continues to be a huge passion in my life. After college, I spent a few years in the UK doing the Part III in Mathematics (equivalent to a master’s degree) at Cambridge.
While I was doing math, I frequently thought about medicine and becoming a doctor. My grandfather was a doctor, and I’ve always admired the humanitarian nature of the work. At the end of my time at Cambridge, I ended up applying to both medical schools and PhD programs in math at the same time.
My thought process was this: Cambridge had some REALLY talented people, math olympians from all around the world. I realized I was never going to win a Fields medal and, at the end of the day, I wasn’t ready to dedicate my life to a career as monastic (albeit beautiful) as pure mathematics. Also, during this time, the Human Genome Project was ramping up. I had a good friend at the Sanger Institute — sort of like the UK version of the Broad — and he got me excited about the potential for applying math to genomics. I made the jump, and went on to medical school.
I did an MD-PhD at Harvard, and my PhD was in biophysics (this a natural home for quantitative people at the med school). The main focus of my PhD research was in applying mathematical methods to analyze genome sequences. On the clinical side, there were a few surprises. When I started medical school, I always thought I would like the more slow-paced and cerebral fields like endocrinology or neurology. As I went through my clinical rotations, I realized I was more of a drama queen! I loved the excitement of cardiology, along with the physiology. I went on to do my residency and fellowship at Brigham and Women’s Hospital in cardiology.
Around that time, a good friend of mine (Dr. Krishna Yeshwant) was one of the founding partners at Google Ventures, which was just getting off the ground. They were interested in investing in genomics, and invited me to be a part of the team. I have to say, this was a really lucky break for me, one that changed my life. Through that role, I came to understand how exciting and dynamic the world of technology is, and it got me thinking about the need for better software in medicine.
After my fellowship, I took a product role at the Broad Institute, where I am now the chief data officer, tasked with building cloud-computing and analytical software for the Institute. My group at the Broad is structured to look much more like a tech company than an academic department. We work hard to create an environment centered around user metrics and products, which may seem strange in a non-profit research institute. However, we’ve all really come to believe in the mission that many of the most important problems in science and medicine need software engineering and data science.
What do you think are some of the most important areas in the health data field at the moment?
Genomics — This is the intellectual odyssey of our day. The quantity of genomic data is doubling every eight months, and the cost of sequencing is falling faster than the cost of computing. This creates legitimate big data problems. The Broad manages about 40 petabytes of data; this requires deep collaborations with various cloud vendors. Scalability is both a great need and a great challenge for the field; achieving it requires savviness with cloud computing and parallelization.
Scalability is both a great need and a great challenge for the [genomics] field; achieving it requires savviness with cloud computing and parallelization.
Medical informatics — There is no problem that has more potential to impact healthcare than improving the quality of our clinical data. Our current electronic medical record systems are a mess — every hospital has different electronic medical record (EMR) schemas, and it’s almost impossible to move data between them. Even when two hospitals have the same EMR vendor it’s still hard to move data between them.
Creating a learning medical system will require cleaning up this mess. If we can do it, we will impact almost every aspect of healthcare, ranging from how we measure the quality of care, to how we insure people, to improving continuity of care in when transitioning between teams. Most importantly, it will enable healthcare professionals to use data to make better decisions.
Fixing these problems will be a heavy lift. It may be painful at first and perhaps is not the “sexiest” area of data science, but in the long term, it’s incredibly high-impact work that could save lives.
If we can [create a learning medical system], we will impact almost every aspect of healthcare, ranging from how we measure the quality of care, to how we insure people, to improving continuity of care in when transitioning between teams.
What is it like to sit at the intersection of tech, the life sciences, and business?
The world of the life sciences is very different from the world of tech. In fact, it’s hard to find two groups that are culturally more different. The world of the life sciences is more hierarchical; tech is more flat. Life sciences companies require a huge amount of upfront capital expenditures, and you need to have a lot of experience to be entrusted with that amount of resource. Software companies can be started with a tiny amount of money, rewarding energy at least as much as experience. The work done by life scientists has the potential to save lives, but with that comes the need for more regulatory oversight. Tech folks want to “move fast and break things.”
We are at a time when these two cultures need to figure out how to work together, as many of the most important problems in science and medicine require software engineers and data scientists to solve them. People who can straddle this divide have a real opportunity. Versatility, egalitarianism, humility, the cultural flexibility to navigate both worlds — these are all necessary skills.
Versatility, egalitarianism, humility, the cultural flexibility to navigate both worlds [of life sciences and tech] — these are all necessary skills.
For example, there are a TON of companies right now that are applying deep learning methods to various areas of medicine, ranging from diagnosing melanomas to reading radiology images. Most of them are founded by luminaries of machine learning, but it’s very hard to find companies that can couple that kind of technical expertise with an understanding of the regulatory environment and complex business models of healthcare.
Part of the problem is that we don’t have a lot of tech and medical professionals talking to each other in a language they can all understand. My dream would be to see a health IT company where the CTO was previously the CTO of Angry Birds, the Chief Medical Officer was a practicing physician that loved to program as a kid, and the CEO was previously the CEO of an ad tech company and were married to a doctor. That team would be invincible! Unfortunately, it doesn’t exist!
What advice do you have for those of us moving from academia into careers in the health data field?
Here are a few of my top pieces of advice:
- Focus on nails, not hammers. What I mean by this is that finding the problems to solve and stating them in a clear way that can be addressed with data scientific methods is often much harder than actually solving them. Building this skill takes time and a deep understanding of the domain. This is somewhat different than an academic environments, which rewards people building beautiful hammers.
- Related to this, it may be tempting to try and apply your quantitative skills to a whole array of different problems at once, as it seems like there is such great opportunity everywhere. But, do commit to something and continue working on it to develop the deep expertise needed to really solve it. This will benefit you in the long run. No one can be an expert in every domain, so find a problem that excites you and take ownership of it by going deep. Along the way, make sure you are always surround yourself with the smartest people that can teach you.
- Become a great coder. It’s important to not just focus on the mathematical methods, but also the software engineering techniques to reduce these ideas to practice. Software engineering is like medicine; it is an apprenticeship model and it takes time and practice, but the rewards are great!
Focus on nails, not hammers.
Commit to something and continue working on it to develop the deep expertise needed to really solve it.
Become a great coder.
Any last thoughts for us today?
I realize that this might be a bit of a scary time for many of you, as you transition from an academic environment to more of a professional one. I definitely remember feeling that way. Keep in mind that, for all of us that are interested in the intersection of healthcare and data sciences, it’s a great time to be alive. It’s literally the intersection of the two most intellectually vibrant fields of our day, and there’s an amazing world of opportunity!
Interested in transitioning to career in health data science? Find out more about the Insight Health Data Science Fellows Program in Boston and Silicon Valley, apply today, or sign up for program updates.
Applications are due March 20, 2017 for the Summer 2017 session.
