Advancing method benchmarking and data sharing through crowd-sourced competitions in cancer research

点击翻译按钮获取中文摘要

Project Summary/Abstract Crowd-sourced competitions have transformed biomedical research by incentivizing the coalescence of communities around timely and difficult problems. These communities have created new standards and benchmarks, and developed innovative solutions to dozens of pressing research problems. The Dialogue on Reverse Engineering Assessment and Methods (DREAM) Challenges is a leader in generating communities through these crowd-sourced competitions. DREAM is an open science, collaborative competition framework that allows participants from around the world to work together to solve fundamental biomedical questions. Its vision is to allow open collaboration amongst individuals and groups, maximizing societal impact through the “wisdom of the crowd”. Over the past 13 years DREAM Challenges have developed key biomarkers and benchmarks for cancer research, fostering the development of innovative methods, the objective assessment of tools and algorithms, the development of community standards, and greater access to critical data sets. To enable the next generation of Challenges, we propose to expand our existing infrastructure into an innovative platform for rapid, rigorous and scientifically-valid Challenge-based assessments. We propose to create a Challenge framework that (1) provides a discovery engine for biomedical Challenges and their outputs, (2) contains innovative tools that enhance and streamline Challenges, and (3) is scaleable and supports distributed assessments using private and sensitive data. These will be married to a suite of educational tools, instructions, best-practices guidelines, and workshops and conferences to expand the solver community and support the organization of independent challenges to maximize impact on biomedical research. Our proposal will utilize several existing and funded driver Challenges in high-impact cancer domains, encompassing key biomedical data in imaging, genomics, EHR (structured and unstructured text), and clinical trial data.