Role of Competitive Challenges in Personalized Medicine

Personalized medicine is widely expected to play a key role in the future of healthcare. It is based on new technologies which enable us to find genetic differences between individuals and as a result give us a chance to identify patients likely to benefit from a treatment. One of the key challenges to its advance is insufficient amount of clinically-annotated genomic data available in the public domain. To address this issue, a consortium named Dialogue for Reverse Engineering Assessments and Methods (DREAM) was setup to bring the larger community to work together on major issues in the field. The main idea is to bring more data in the public domain and to pose personalized medicine problems as competitive challenges (competitions) with deadlines and non-commercial rewards for winners. Anybody from around the world is allowed to participate, and researchers from leading world universities have joined in. In the latest challenge the goal was to predict relevant clinical outcomes of Acute Myeloid Leukemia patients treated with standard chemotherapy regimen, using a combination of clinical (age, sex, lab test results etc.) and genomic data. The final results of the competition were released in October 2014. Team Clinical Persona placed third in predicting remission duration and fourth in estimating overall survival. The purpose of our talk is twofold. First, we show the benefits of public competitions in science and especially in predictive modeling. Second, we show our lessons from participating in the challenge.