Originally written by YAN KOU from Insight Health Data: The Emerging Field of Health Data Science in Boston
What if you could pick out early warning signs of heart conditions out of somebody's Fitbit data? It turns out that you can.
This technology was developed by Mike Klein, a neuroscience Ph.D out of McGill University, as part of the Insight Health Data Science Fellowship. Insight offers a Fellowship three times a year where academics learn the applied data science skills they need to work in industry. Klein had already used machine learning in his research - it comes in handy when interpreting fMRI data - but at Insight he picked up industry standard tools and data science workflows to handle messy datasets. As a long-time wearable enthusiast, Klein was fascinated by the Health eHeart study at UCSF - where over 30,000 participants contributed Fitbit step data and anonymized clinical information. After many hours of data wrangling and feature engineering, Mike was able to detect early warning signs of heart conditions based on Fitbit walking patterns.
"I was excited to be able to leverage my existing knowledge of machine learning techniques to attack a problem in a new domain area with very different sorts of data. Both the industry focused tools and the pace (a sprint!) were very different at Insight and, in the process, I learned a great deal about working with sparse/unbalanced datasets." says Mike Klein
Klein wasn't the only one in his cohort to come up with a data-driven solution during the intensive 7-week fellowship. Greg Koytiger, a former postdoc in systems biology at Harvard Medical School, was interested in precision medicine on cancer therapeutics. "The motivation is to save time and cost on finding the right patients for the right treatment", Koytiger said. He built a series of models trained with gene expression profiles of human cancer cell lines to predict the link between genome-wide expression and drug sensitivity. By applying his models to breast cancer patients treated with Docetaxel, Koytiger correctly identified 91% of responders and 85% of non-responders.
"Predicting drug response will help oncologists identify the right drug for their patients and, help pharmaceutical companies design targeted proof-of-concept trials for experimental therapeutics. Using sparse linear models allowed me to understand the genes responsible for drug resistance and sensitivity thereby shedding light on potentially interesting new biology." says Greg Koytiger, who recently signed an offer with Immuneering after completing the Health Data Science program.
These stories are just a snapshop of the emerging field of health data science. Across the health and life sciences sector, servers are overflowing with data: electronic medical records, claims data, genome sequences, drug screens, clinical trials records, longitudinal studies, and even quantified-self data and more. This information is a form of "big data" not only in sheer size but also in complexity, diversity, and chronicity. Health data scientists apply quantitative computational approaches to understand health data. We are starting to see how health data scientists could improve the efficiency of clinical trials, personalize health recommendations, build new tools for physicians, consumers, insurers and regulators, to name a few.
"Applying big data strategies to better inform decision making could generate up to $100 billion in value annually across the US healthcare system." write the authors of Big Data: The Next Frontier for Innovation, Competition, and Productivity, a comprehensive research study published in 2011 by the McKinsey Global Institute.
The challenge that health organizations face is to find data scientists with applied analytical training, and more importantly, an understanding of the problems in healthcare. The mission of the Insight Health Data Science program is to train academics to develop the applied data science skills required to meet these challenges. This summer, Novartis along with The Broad Institute, Memorial Sloan Kettering Cancer Center, Athenahealth, Tamr, Seven Bridges Genomics, and many fast-growing health startups joined Insight for its inaugural Health Data Science Fellows Program.
Novartis' Executive Director of Informatics Systems, Stephen Cleaver, commented "Insight gives people a practical path to move into informatics".
Insight trains future health data scientists through an intensive 7-week Fellowship program in which Fellows build data products under the supervision of industry mentors. Insight Fellows come from world-class research programs with years of experience in data and code. Insight creates a collaborative learning environment where physicists, neuroscientists, statisticians, computational biologists, quantitative psychologists, biomedical engineers from Harvard, MIT, and other top schools bring their expertise and work together daily to build health-related products, using cutting-edge machine learning techniques and industry-relevant tools such as python, SQL, and Amazon Web Services. "Insight Fellows can have an outsized impact in the world by combining data science best practices from across various industries and applying them to healthcare and the life sciences," says Jake Klamka, Insight's founder.
During the first Health Data Science session, Insight Fellows worked on a variety of projects using rich public health data from dedicated databases, mining from unstructured social media data among other resources. To overcome the limited availability of sensitive, healthcare data, Insight brings industry and research partners together to identify real-world problems for Fellows to explore. Partners including UCSF Medical Center, Sage Bionetworks, Comprehend and One Codex.
During the Health Data Science program, industry mentors came to Insight to share their perspectives on the emerging intersection of healthcare and data science. In the inaugural session, Fellows met with Anthony Philippakis, Venture Partner at Google Venture; Andy Palmer, CEO at Tamr; Zen Chu, faculty advisor at MIT Sloan School of Management; Martin Leach, Vice President of R&D IT at Alexion Pharmaceuticals and Paul Nagy, co-director of the Johns Hopkins Medicine Technology Innovation Center and many others.
In addition, Insight engages its alumni community to provide mentorship for the current Health Data Science Fellows. Insight has been running for over 3 years, with programs in Silicon Valley, New York City, and Boston. Over 400 Insight alumni now work on leading data science and data engineering teams at Facebook, LinkedIn, Netflix, Uber, Airbnb, Apple, The New York Times, Bloomberg, Khan Academy, Invitae, 23andme, and many other top companies.
The unique blend of thought-leaders, industry data teams and alumni provides Insight Fellows with an effective way of learning the opportunities and challenges in health data science. As Adam Jenkins, Insight alumnus and now a data scientist at Biogen, summarized:
"Insight provided me with the opportunity to connect with industry leaders and to learn how professional data scientists shape companies. Through onsite company visits and mentoring with past Insight fellows who now hold industry positions, I was able to gain a firm grasp on the landscape a data scientist works within that I was unable to learn from an academic background."
The first cohort of the Insight Health Data Science Fellows program are currently interviewing with various top data science teams in Boston. The next Insight Health Data Science Fellows Program will start in January 2016. As Yan Kou, Program Director leading the Insight Health Data Science Fellows Program in Boston puts it, "Our aim is to enable the future data science leaders in healthcare."
Find out more about the Insight Health Data Science Fellows Program.
Companies and organizations interested in partnering with Insight can email info@insighthealthdata.com
