Big Data and Personalized Medicine

Steven Friend is being honored as a Champion of Change for the vision he has demonstrated and for his commitment to open science.

The massive public investment in the Human Genome Project has already paid serious dividends - calculated by some at $141 in returns for every $1 invested. But the advances in genomics have yet to impact the lives of most American citizens through what is often called "personalized medicine.”

One promise of personalized medicine is to predict which patients will respond to medications and which patients will not. From molecular data like our DNA sequence information, we should be able to tease out the subtle variations that make each of us unique to predict, for example, whether a drug is likely to work in our bodies. But so far we haven’t been able to do this with real effectiveness, or at the kind of scale we need. That means many people receive drugs that are unlikely to work, for reasons that we should be able to understand - but we don’t.

Rheumatoid arthritis (RA) is a good example. For people with RA, strong immunosuppressive medications are administered in order to treat pain and inflammation.  Despite substantial efforts by researchers in academia and industry, there are no reliable genetic clues to predict which 30% of patients will enter clinical remission - that is, have the drug actually work - following treatment with therapy in RA. That means we have to give the drug to everyone, and pay for it for everyone, while 70% of Americans who take the drug only get the side effects without any therapeutic relief.

Alzheimer’s disease is another area where we haven’t been able to use genetics as well as we should. We have amazing imaging equipment that can peer deep into brains, and we have amazing sequencing capacity. But we haven’t yet put them together to figure out why some people get Alzheimer’s and some don’t, or why some who get the disease progress faster than others. Some barriers have been created as a result of the small numbers of those who know how to use the complex information, as well as from fact that so much data sits compartmentalized inside corporate and academic silos that have limited appeal and accessibility for scientific collaboration.

The time is right to try a different approach to make genetics work for personalized medicine.   At Sage Bionetworks, we are partnering with innovative groups such as Gustavo Stolovitzky at IBM-DREAM, the Robert Wood Johnson Foundation, and Ashoka to apply the tools of open science to solve big problems in health.

For example, we’re launching Big Data “Challenges” in both Rheumatoid Arthritis and Alzheimer’s disease that have the potential to bring the power of crowds to figure out how individual genetic variation impacts these diseases. We’re making the Challenge data about genes and proteins available to anyone who wants to take a crack at the problem.  To drive these Challenges, we’re leveraging our open-source Synapse data-sharing platform and DREAM’s well-established framework for running Challenges (originally developed by IBM).  DREAM’s know-how helps us design smart, impactful Challenges, and Synapse’s leaderboards, code-sharing and provenance tools will get teams of teams revved up and participating in a real-time dialog that fosters rapid learning and better predictive models.   These Challenges will generate winning models that then guide new clinical trials (in RA) and that spell out the patient data we most need to help guide better treatment for patients (in AD). 

By partnering with the Arthritis Foundation, the Global CEO Initiative on Alzheimer’s Disease, and others, we know we can make an immediate impact on these two diseases.  And most importantly, we can do it in a way that others can build on - which is the essence and value of open science.

Steven Friend is the President of Sage Bionetworks.