Figure 2: Schematic of the Pan-Cancer analysis workflow.

Data were aggregated and standardized from the TCGA DCC, Broad Firehose and individual analysis working groups and processed into easy-to-use tab-delimited files. Collaborators used a variety of analytic tools, such as R, Python, Unix shell and the web client, to interact with data in Synapse while also storing results, provenance records, analysis descriptions and source code. For a subset of these results (for example, patient survival predictions), Synapse carried out automated performance evaluations and displayed results on a real-time leader board, which were available to collaborators to perform comparative meta-analysis or adapt model source code to additional applications.