AstraZeneca ($AZN) has opened up a wealth of preclinical cancer data and challenged people to dig into it to predict the effectiveness of different drug combinations. The data dump covers 11,500 combinations of 118 drugs across 85 cancer cell lines that AstraZeneca tested for cell viability.
In addition to this resource, which the organizers of the challenge think is the largest publicly available combination screen, participants have access to gene expression, somatic mutation, copy-number alterations and methylation data for 85 cell lines. Sanger Institute is contributing the genomic data. The idea behind the release of the data is to provide researchers from any organization with the building blocks of models to predict in vitro drug synergy. Having such models could take some of the guesswork out of the increasingly critical task of creating effective cancer combination therapies.
AstraZeneca and its collaborators have multiple aspirations for the projects. The two core challenges for the crowd are to predict in vitro synergy results, both with and without the use of experimental data. Restricting the use of experimental synergy scores mimics the conditions faced in real-world R&D and as such could lead to a useful model. The collaborators also hope to uncover fundamental traits that characterize effective combinations and biomarkers that can predict synergies. Teams can start submitting models and receiving feedback from the organizers from mid-October onward.
The project is being organized as part of the DREAM Challenge, the nonprofit open innovation group that has spent almost the past 10 years encouraging crowdsourced research. AstraZeneca is the latest in a long line of organizations to be attracted by the idea. The Broad Institute and MD Anderson Cancer Center are among the organizations to contribute data in the past, while Google ($GOOG) and Sage Bionetworks have both provided computing resources. Research groups at Roche's ($RHHBY) Bina Technologies are among the teams to have entered the challenges.