Postpublication data sharing: the Rome Agenda
Sarah Tomlin
Wednesday, 02 September 2009 20:39 UTC
Sharing knowledge is key to the progress of science, but researchers do not always release data and research materials, even after publication. Earlier this year, two research communities held meetings with a broad range of stakeholders to discuss ways to promote data sharing in biology, and they report their recommendations in two Opinion articles in this week’s Nature. See special issue on data sharing here.
Mouse researchers, along with funding agencies and publishers, met in Rome in May to address the barriers preventing more effective sharing of data and biomaterials — particularly mouse strains and embryonic stem cells. Their agenda, free to access here, suggests guidelines to enable sharing of materials under the least restrictive terms, avoiding material transfer agreements where possible.
The Rome participants argue that funding organizations, journals and researchers need to work together to encourage better use of public repositories and to promote a ‘research commons’ in mouse biology.
The recommendations are intended to spark community discussion on this subject. Paul Schofield and others will be responding to reader comments in this forum. Be sure to have your say.
Updated 09 September 2009 17:05 UTC
-
Replies
-
A small comment: This conference was focussed on mouse genomics, but as an attendee who is involved in geophysics, I think it is important to note that the principles laid out in the Rome Agenda extend well beyond genomics and even life sciences. Indeed they are fundamental principles for all of science.
-
what about sharing code instead of data? has there been a special issue or a conference about that?
-
I think that smart sharing of the data will be a good move for society. The more scientist can share the data about his work the more he can earn credit for it.
-
Chris, the meeting did discuss the sharing of informatics tools – ie code – at the same time but we didnt focus on this in the opinion piece. There was universal agreement that the code for tools should be shared on the same basis as bioresources like mice and ES cells. The Informatics community has gone some way to facilitating this with open standards such as GPL, but the sharing principle applies to this area as much as it applies to mice.
We dont plan to have a specific conference on this – we recently met to discuss database and resource sustainability which was the last meeting we planned in this series. A lot of the conclusions from this meeting were reported in the Nature Editorial on Thursday 19th Nov.
Maybe time for someone to look at the particular issues of code sharing?
-
How come that researchers do not always release data and research materials, even after their work has been published?
-
Well there are lots of reasons, some of which are discussed in these two articles, and in many other articles.
-
Tim, I think that there are two or three predominant reasons. Numbers one and two are probably a combination of neglect and, often, the technical inability to submit complex data sets to public repositories. Once a paper is published then most scientists see this as the culmination of a body of work and the end of a set of tasks. To then go about formatting and submitting data to public repositories is seen as a lot of extra work without any real immediate gain unless journals insist that it be done before publication. Often there is also a manpower issue as well – what is the best use of my postdoc’s time? Secondly submission of complex datasets can require some technical expertise , for example in formatting data or adding metadata and many laboratories just doent have this expertise or any informatics backup. Of course in some cases there are not appropriate repositories for some data types and in this case sharing relies on the originating lab or a Journals Supplementary information site.
The other main barrier is simply fear of helping the opposition in what is often highly competitive science with research funding and employment issues, paramount in making the decision whether or not to share large datasets which might contain useful information for competing groups.
Finally, what I hope is the smallest group – those who have some concerns that their data might not be reproducible or may have been incorrectly analysed. Whilst not necessarily implying fraudulent intent, worry about how well the primary data was technically analysed can be a disincentve to sharing. Of course it is down to jornals and referees to insist on the availability of primary data during the revoew process, but its not often that referees actually try to replicate the analysis of the primary data before a paper is accepted.
-
