It’s Up to You to Start the Company
As the founder of three different data-driven companies, Atul Butte, Director of the UCSF Institute for Computational Health Sciences knows a thing or two about biotech entrepreneurship.
Genomics abounds with commercial opportunities, new ideas and technologies just waiting to be licensed and developed into the next world-beating research of healthcare product. More and more, a solid entrepreneurial skillset will come in just as useful as a dab hand in the laboratory. In fact, according to Atul Butte, Director of the UCSF Institute for Computation Health Sciences and founder of no less than three different biotech companies, there’s very little difference between writing a research grant, and writing a business plan, and the latter might just make the benefits of all that data available to the wider public.
FLG: You’re a rare breed in that you have remarkable experience in computational and medical fields at some very impressive organisations. I was wondering if you could talk a little about what got you interested in computer science in the first place, and how your career has taken you up to where you are now?
AB: Sure, I’d love to. When I was growing up, the personal computer revolution took off. It just exploded in the United States. So my first computer that I started to learn how to program with was an Apple II+ that my parents got me. Prior to that, I was writing programs, so I’d take that to the department stores to plug in to the computer there. So they obviously figured out that having a regular computer for the home, which was ridiculously expensive, but obviously now, in hind sight, was really worth it.
I knew I wanted to do computers and medicine, so I was volunteering at hospitals and things like that, that high school kids do. At the time National Geographic magazine had these covers of ‘The New Radiology’ – MRI and CAT scan. So I thought for the longest time if I wanted to do computers and medicine, I was going to be a radiologist. I went into a program at Brown University in Providence, Rhode Island. They have an eight-year program, where you major in anything that you want and you’re guaranteed to go into the medical school if you keep your grades up. So that’s how I was able to do computer science, and still go to medical school there. So I got the full computer science experience; I spent one winter at Apple, writing software; I spent some time at Microsoft on the Excel team; so I learned how to write code. I went on to medical school and took a year off, and during that year, the Human Genome Project started up. To me, that was the ultimate in digitising biology and medicine, and I never looked back at radiology from there! I really got enamoured by molecular biology and genomics, and after medical school I went straight into residency and training in paediatrics. My mentors convinced me I should go back to MIT and get my PhD, so I did that while I was seeing patients in Boston. That’s probably the only place you can see patients and get a PhD at the same time. Then in 2005, I moved to Stanford and worked there for 10 years. Last year, I moved to UCSF where I’m the new
FLG: When people hear your name, one of the first words that come to mind is ‘data’. How is data changing the way we approach scientific research?
AB: We are able to measure so many things now, right? In genomes, we have gene expression microarrays, that’s how I got my start, those gene chips from Affymetrix; we have proteomics, cell counters, cell sorters, all of these technologies that can measure so much now. I think that has really changed a lot. In the beginning, there was a lot of resistance to this kind of technology. I mean, there were derogatory terms like ‘fishing expedition’, things like that where, in the old days, people thought you should just have one idea and just chase it down, and if it doesn’t work it doesn’t work. And now you really survey, you scan, you screen, the entire genome, the entire proteome. The other revolution that happened is not just that we can make these measurements, but the measurements are now increasingly open to the public. Because of transparency, and increasingly now because of reproducibility, we’ve got all these pushes to get the data open to the public. So many people make these measurements, and so many people make them public, but so few people use that data. To me, that’s really the ultimate treasure, it’s all this raw scientific data from these top scientists, who get a lot of funding to do this. Yet so, few people are actually using that data, so it seemed like a natural kind of fit to my background.
FLG: Does this emphasise the need for greater standardisation in science?
AB: It’s not a standards problem. There are not that many different standards. If you really want to use the data, you’ll learn how to use it. It’s not super sophisticated. Yeah, there might be a few databases out there when you get started, but there aren’t 500 of them, right? Each one has a standard way of representing it. We do also have standard tools like R that we use for data sciences. I think the real challenge we’re facing is not standards, but the lack of training. We don’t have enough people in this field, period. There are so few people trained in biomedicine and in anything computational, the quantitative sciences. You need both backgrounds. You could go pretty far being a computer scientist, but you could get much further if you know what questions to ask. If you know the questions that everyone has been dying to answer, and you have the ability to answer them, that’s how you’ll go farther. So that’s not really a standards problem. The real problem is there aren’t enough people trained in a way that lets them actually use all the data that’s out there.
FLG: A lot of focus of the Precision Medicine Initiative (PMI) has been on the genomic aspect, but it’s just one part of the project. How big a difference will it make to be able to use Electronic Health Record data and environmental data in conjunction with genomic data for scientific research?
AB: Yeah, that’s a great question. So specifically, we’re talking about the PMI, the federal initiative by NIH and the White House, where the aim is to get a million volunteers, and put all these data together and ask and answer interesting questions. So right now, they’re in the middle of figuring out the funding, and who’s going to get the funding to actually launch or manage the project. So all that’s underway right now. They’re guessing that half the participants will come from the healthy public and half will come as provided by health provider organisations.
In general, the ones that come from the healthcare providers obviously have some medical history, or else we, the hospitals, wouldn’t have known that. So those are the ones whose medical records are probably going to be very interesting. For the lay public, which is actually really healthy, it’s kind of rare to have much of an electronic medical record. If you’re essentially healthy, there’s not actually that much in there, maybe some random blood tests, so there’s less useful information there. Obviously, there is this focus on genomics, but it’s not really obviously stated that we’ll be getting genetics on these people. We’re guessing that’s what they’re going to do, and that’s a strong guess. But there are also efforts to try to get behavioural measurements, so smartphones, mobile devices, wearables. So I think there’s going to be a large set of data that very few people know how to analyse, especially in academia, especially for population management, and I think that’s going to be a new direction beyond just genomics.
You can read the full interview in the 9th issue of Front Line Genomics Magazine.