Return to 2012 Open Space/Unconference discussions

Synopsis

In the traditional research pipeline utilizing modeling and simulation framework, data/software/models are commonly established by a handful of investigators and disseminated afterwards. How can open development, crowd-sourcing, and cloud computing can be integrated for community-driven modeling & simulation and computational medicine? How can we overcome constraints associated with traditional academics to facilitate this?

Scribe

Ahmet Erdemir, erdemira@ccf.org

Participants

• Roger Mark
• George Moody
• Scott Christley
• Zahara Sotoudeh
• Yuekai Sun
• Tony Macula
• Jacob Barhak
• Chase Cockrell
• Stacey Finley
• Ahmet Erdemir

Notes

• Discussions started with delineation of dissemination (after research is done) and open development practice (interactive community input during research is being conducted). The potential advantages of open development practices were summarized; distribution of workload to diverse groups, benefit for multiple people being able to do quality check, improvement of procedures, etc. Potential constraints for open development, particularly related to its acceptence in traditional academics were listed; conflict with 'publish' or 'perish' mentality, funding and award mechanisms appreciative of open development efforts, sharing of intellectual property (particularly when home institutes of participants are considered). Examples for necessary infrastructure for collaboration infrastructure (https://simtk.org), for computing infrastructure (http://www.xsede.org), and for project specific interfaces to engage contributors (http://robetta.bakerlab.org, http://fold.it/portal) were provided.

• It is realized that general purpose development and computing infrastructures exist.

• Some examples for open development, including semi-crowd sourcing, were given; http://www.innocentive.com, http://www.netflixprize.com, http://www.heritagehealthprize.com

• Open development was seen as a further push expanding upon post-research model/data/resource sharing, which also indicated current problems to encourage dissemination and establish it as the norm. Model/Data sharing practice still remains problematic. One participant proposed asking patients to obtain direct permission to use their data, removing all the players in the middle. Another noted the role of dissemination for establishing reproducibility.

• Challenges to engage experts to contribute were noted; and the likelihood of recruting less experienced but more eager contributors was realized. This also indicated the need to frame the crowd-sourcing and open development procedures in a way that is accessible by the less experienced and that can allow useful contributions from them.

Action Items

• Inform institutes of participants and funding agency to promote acceptance of model/data/resource dissemination efforts in an equivalent manner with publication efforts (knowledge dissemination)

• Identify specific challenges for community driven research, which can be shaped to be solved by novice crowds, e.g. sensitivity analysis, model/data documentation, reproducibility and benchmark testing, human annotation of data

• Implement a given challenge to conduct a social experiment