Julie Steele

Strata Rx Online Conference: Personalized Medicine

Date: This event took place live on October 05 2012

Presented by: Julie Steele

Duration: Approximately 120 minutes.

Cost: Free

Strata Rx brings together experts and innovators in data science and healthcare, crossing traditional industry boundaries and uncovering insights on the emerging big data and analytic approaches that can help solve some of healthcare's biggest challenges.

In this free online conference we will discuss how Microsoft Research has developed a new version of the Linear Mixed Model algorithm that is not only computationally inexpensive, but also is better at finding the true signals that account statistically for the diversity of people in data sets used to identify "complex traits".

You will see how using crowd-sourcing and distributed distribution models can create a massive phenotypic data set of the human condition to compare to genetic data sets so we can better understand what an individual genome represents.

We will also discuss the privacy implications of personalized medicine: potential pit-falls and policy decisions that we're likely to face as our analytics improve.

Hosted by Strata Rx co-chair Julie Steele

Julie Steele is an editor at O'Reilly Media interested in connecting people and ideas. She finds beauty in discovering new ways to understand complex systems, and so enjoys topics related to gathering, storing, analyzing, and visualizing data. She holds a Master's degree in Political Science (International Relations) from Rutgers University.

Julie also works with topics related to the languages Python, PHP and SQL, and is co-founder of a group of non-programmers learning Python. Julie lives in NYC where she eats, reads, codes, and practices yoga.


Discovering Genetic Associations on Large Data
David Heckerman

One of the key components of personalized medicine is the discovery of relationships between the genome and medically relevant traits, such as the propensity for a given disease or how well a particular drug will work. Early on, it was believed that small numbers of genetic markers would have a strong effect on a trait. Now, however, it appears that many traits are influenced ever so slightly by a large number of genetic markers. To uncover these subtle relationships, large amounts of data are needed. With large amounts of data comes confounding, for example, due to data coming from closely related individuals or individuals across multiple ethnicities. Such confounding leads to seemingly real yet false results. Until recently, models that could correct for such confounding were too computationally demanding to be used on large data. In my talk, I will describe how these models can be made fast without sacrificing accuracy, allowing application to data covering hundreds of thousands of individuals.

About David Heckerman

David Heckerman is Senior Director of the eScience Group at Microsoft Research. Since 1992, he has been a researcher at Microsoft, where he has created applications including the first content-based spam filter and web services for medical diagnosis. His research is in the areas of statistics, machine learning, and artificial intelligence with applications in medical diagnosis, the design of a vaccine for HIV, and the search for genetic causes of disease. He received his Ph.D.(1990) and M.D.(1992) from Stanford University. His Ph.D. dissertation on automated medical diagnosis received the ACM doctoral dissertation award. David is a AAAI Fellow and a Distinguished Scientist at Microsoft.


Crowd-Sourcing Phenotype Data
Michael Simpson

In this talk Michael Simpson discusses crowd-sourcing phenotype data of the human condition and how this data is used to investigate the unexpected correlations between our health, our unique distinguishing traits, and our lifestyles. Don't miss this fascinating discussion.

About Michael Simpson

Michael has 20 years experience in executive management, consulting, and product management. His management experience spans from large-scale data aggregation to web services. Michael received a BS from Cornell University and an MBA in finance from the University of Texas at Austin.


Advanced Analytics for All: Enabling business users to act on length of stay patterns at a leading hospital system
Arijit Sengupta

The future of analytics is simplicity, ubiquity, and actionability; in other words, - Advanced Analytics for All (A 3). In an A 3 approach, the focus is on users without Statistics or Computer Science skills collaborating to leverage Advanced Analytics. This session will cover an anonymized case study on a specific A 3 project at a leading US hospital system that analyzed what factors were the best predictors of patient outcomes.

About Arijit Sengupta

Arijit Sengupta is the CEO of BeyondCore, the Chair of the Big Data and Advanced Analytics SIG at the Service Research and Innovation Institute (SRII), the Chair of the Cloud Computing Chapter of the International Association of Outsourcing Professionals (IAOP), and a member of the Cyber Security Advisory Group set up by NASSCOM and the Government of India. BeyondCore received awards such as Gartner Cool Vendor in Business Process Services 2012, GigaOm Structure Launchpad 2011 and the UP2010 Overall Most Innovative Cloud Provider. BeyondCore Lucid automates the complex statistical analysis currently conducted by human analysts and presents the key business insights in a manner that a business user than easily leverage.

Arijit has guest lectured at Stanford and other universities; spoken at conferences in a dozen countries; and was written about in The World Is Flat release 3.0, the New York Times, San Jose Mercury News, and other leading publications. Arijit held leadership positions at several eBusiness initiatives and previously worked at Oracle, Microsoft, and Yankee Group. He has been granted eight patents in the domains of advanced analytics, Business Process as a Service, business process improvement, operational risk, privacy and information security. Arijit holds an MBA with Distinction from the Harvard Business School and Bachelor degrees with Distinction in Computer Science and Economics from Stanford University.


Overview of Privacy Concerns and Regulatory Challenges Concerning Personalized Medicine — and Some Modest Suggestions for Change
Ann Waldo

This talk will outline major concerns and questions regarding privacy and ethics in the rapidly developing personalized medicine field. In addition to pondering how technology is outpacing law, we will focus on specific privacy legal provisions that pose barriers to genetic research and data utility today. We will offer some specific and practical suggestions for how regulatory obstacles to personalized medicine research could be eased.

About Ann Waldo

Ann Waldo is a partner in the boutique law and public policy firm of Wittie, Letsche & Waldo, LLP in Washington, DC. She graduated from the University of North Carolina School of Law with high honors and is a Certified Information Privacy Professional. Her law practice focuses on consumer and health-related privacy, information security, and health care reform issues. She also handles public policy, external relations, and government relations regarding privacy and health care. She has particular interest in emerging technologies that handle sensitive health information, such as personal health records, health data companies, social and mobile health, and health information exchanges.

Ann was previously the global Chief Privacy Officer for Lenovo and the Chief Privacy Officer for Hoffmann-La Roche. She worked in public policy for GlaxoSmithKline and served as in-house counsel at IBM, working on consumer protection, marketing, and e-business. She is active in the International Association of Privacy Professionals, has consulted with foreign governments regarding privacy laws, and has represented the United States government in APEC privacy talks in Korea and Australia. She serves on the Board of Advisors for the Harvard SHARP grant on substitutable electronic health record components.

Questions? Please send email to