Public Access to data and models: Yes No Maybe

As I mentioned in my last post about public access to data and models in context of IUPS Physiome project and the Virtual Physiological Human (VPH) framework, there was some serious debate on this agenda during VPH Community satellite symposium of IUPS 2009. Professor James Bassingthwaighte from University of Washington, who also leads the NSR Physiome project suggested that public access to data and models is impractical and counterproductive (arguments were presented by Peter Kohl on behalf of James). It is important to note that James and Peter played the role of devil advocate during this discussion and opinions expressed during this discussion do not necessarily reflect their personal views. According to James, focus of modelling community should be on redevelopment and not on reproducibility, reproducibility is somehow overrated. Redevelopment is the best way to penetrate a model or hypothesis, and better ideas will emerge when other investigators will attack a model or hypothesis from scratch and attempt to improve on it. Redevelopment supports removal of inconsistencies within data or model and sparks alternative model development. Sharing may cause misuse & hype – opportunistic followers will generate hype while putting a model into the hands of naïve users invites misuse. Furthers he think that sharing will raise the expectations and by sharing you commit to support efforts of others leading to more responsibilities and extra workload. In his words model sharing can be suicidal for careers progression as well as for future opportunities. Ethos versus reality-sharing and caring is fine but Why me at first place? In my opinion much of issues raised by James are genuinely correct, and they reflect a general fear in scientific community about open access to data and model. Few are worried about losing competitive advantage while others just don’t want to bear the burden of revealing their work to others. Does open access to data and model hurts the potential scientific collaborations? For example we generally see complimentary efforts from wet lab and computational lab all together but in case of open accessible data and model, one will not need other. Modelling community needs a bigger picture on the model and data sharing issues and that will be possible only if we raise the questions. Later Marco Viceconti presented a point by point counter argument that open access to model and data is scientific necessity and moral obligation. Marco suggests that reproducibility is a fundamental tenet of science and sharing is the only way to conquer the systematic falsification. Once we accept that model and data sharing is mandatory then research systems will align to it, and technology will make model and data sharing more easier and less time consuming as much as it has simplified scientific publishing. As scientists we are expected to share with humanity the new knowledge we produce, in a way that make scrutiny of possible falsification while at same time it ensure accurate replication. Ideal research is only about collaboration, and modern research practice oscillates between collaborative competition and competitive collaboration. What we need is to community level competition and not individual pursuit. Domain those share their knowledge and data will advance fast which means they will have more funding and opportunities. Marco also stressed that we must change out reward system to acknowledge model and data sharing, which will require to go beyond the impact factor game. Currently rewarding and funding systems is based on quality of peer reviewed publication which is usually measured in terms of different citation matrices such as impact factor and others. Can we track down the data and model reuse to create a scoring system which will encourage the sharing of data and models. Technically it is a feasible idea, for example Michael Hucka suggested that shared model can have public key encryption where original creator can track the model reuse flow. Andrew McCulloch cited a particular example of Protein Data Bank (PDB) especially how crystallography community and associated journals have encouraged the pre-publication submission of experimentally determined protein stricture in PDB database. But I am surprised how many structures reported in PDB are reproducible, I guess no one ever tried to do so although multiple isoforms or variants of a protein are reported to have a common structural trend. In between James Lawson from Auckland Bioengineering Institute made an interesting suggestion that in future we should adopt a two tier publishing systems- traditional publishing of research in peer reviewed journals, and publication or sharing of model and data. In summary whole debate was interesting but I felt the modelling community is quite unaware about recent example of data sharing in other domain such as open notebook science (ONS) solubility effort which allows immediate communication of scientific results. Also the role of funding agencies is crucial, NIH and other agencies such as Welcome trust have already established a positive mandate for open access publications, a similar kind of approach is needed for open accessible data and models particularity if research is funded by tax payers money.

Note: This article is based on my notes taken during the VPH symposium. Unless explicitly stated in the post, everything written by me is my personal opinion and it does not reflect the opinions of my employers. The content and opinions expressed in this article are not guaranteed to be correct, I could be wrong. All I can do is make corrections on a simple request.