The Metadata Automation DREAM Challenge has closed! The winners are C-Path Analytics Team in first place and tied for second are ENGRDynamics and the CEDAR Team. 

Supported by funding through the Cancer Moonshot^sm, this challenge asked participants to develop algorithms to automatically annotate data fields and values from Applied Proteogenomics Organizational Learning Outcomes (APOLLO) network and The Cancer Imaging Archive data sets. Solutions for automating the annotation will play a critical role in allowing researchers to search for data and make discoveries across different collections, formats, and repositories, including the NCI Cancer Research Data Commons. To validate and refine their algorithms against NCI data and known data standards, each team submitted a test file for each phase of the challenge. In the final phase, each team was scored on how well the algorithm:

•  matched the data elements to the specified annotation gold standard.
•  overlapped between the identifies in the data set and the annotated gold standard.
•  covered between the submitted values and annotated observed values.

The challenge may be over, but the work to automate metadata for repositories will continue. As an extension of this challenge, the three winners will collaborate with two additional teams from the challenge, meta_n00bs and pb2pv, to improve algorithms by presenting their approaches with each other. After the collaboration has ended, the teams will draft a white paper sharing their solutions and possible future approaches.

Congratulations to the winners and thank you to all those who participated in this challenge.

Learn more about the winners and their proposed solutions.