Open access is not enough on its own – data must be free too

If your research has been funded by the taxpayer, there's a good chance you'll be encouraged to publish your results on an open access basis – free at point of publication and with reuse and redistribution rights.

This final article makes publicly available the hypotheses, interpretations and conclusions of your research. But what about the data that led you to those results and conclusions? Isn't the underlying data just as important to support the quality of the findings?

A huge amount of data is being produced by scientists every day, but too often key information is left to rot in an Excel document on someone's desktop, or handwritten in a notepad that is later thrown away.

Increasingly, policymakers and funders are introducing data-sharing and stewardship policies to solve this problem. Funders want to see this data being properly described, stored, shared and reused, to realise its full potential. Data producers are also somebody else's data users, and they have also come to the same realisation. Open data ensures that the scientific process is transparent, helps others to reproduce results and can even help speed up the process of scientific discovery.

Open data isn't always easy to access

In practice, it can be intimidating to scientists who might be afraid of being told they've made a mistake in the data collection process. It can also be a thankless task as it takes time, and researchers who do a good job are not necessarily rewarded. Researchers may be concerned about others finding discoveries in their data, before they themselves have had a chance to exhaust its potential.

Even when data is made available, the detail given isn't enough to really understand how an experiment was conducted and the results produced.

To encourage researchers to think about how they need to manage and share the data they produce, most funders' data-sharing policies ask the researcher to write a data management plan – but these aren't scored, so in theory you could write a vague one and still get a research grant approved. Nevertheless, more stringent data-sharing monitoring policies are being developed.

So data that is in theory open and free to access may still be hard to get hold of. It may not have been stored or cited in the appropriate manner; it may not be interoperable with related data because it is not formatted appropriately; or it may not be reusable because it may not contain enough information for others to understand it.

To combat this trend, a new type of publication has emerged – the data paper, which describes a particular dataset or a group of datasets, and is published as a peer-reviewed article in a scholarly journal. Data papers provide a missing link between the data and the research article. But most of the publications have a very narrow focus or are still narrative-based, with an emphasis on interpretation and conclusion.

Data descriptors

Earlier this month, Nature Publishing Group launched Scientific Data – a broader, interdisciplinary publication dedicated to a more specific type of data paper: the data descriptor. This new category of peer-reviewed publication provides detailed descriptions of individual or combined experimental, observational and computational datasets. Data descriptors include a narrative article (human-readable) accompanied by structured information (machine-readable), annotated by the in-house editorial curation team.

Data descriptors are designed to contain all the information required to allow people to find, interpret and reuse the dataset. Because they are peer-reviewed and citable, they enable researchers to get credit for their work and incentivise them to make their data more discoverable.

A data descriptor doesn't contain tests of new scientific hypotheses, analyses providing new scientific insights, or descriptions of fundamentally new scientific methods. Instead it contains the steps required to understand and reproduce the experimental steps data and link the resulting data files and their location.

Scientific Data accepts publications that link to data stored in subject-specific community repository, or in a general-purpose repository such as figshare or Dryad. Lists of recommended repository and useful standards – to enrich the description of structured information – are being developed with community-driven efforts such as BioSharing and ISA.

The work on improving access to and reuse of data across the board continues. Efforts by the Research Councils UK policy and the Royal Society report are UK exemplars. Across the pond, the NIH Big Data to Knowledge Initiative is developing a data discovery index for biomedical research.

This is a good start, but the scholarly community, from researchers in both academia and commercial arenas, to librarians, publishers and funders, need to do more to transform research and publishing in data-centric enterprises for the science of the future. And it's not going to happen overnight.

If you'd like to be part of the conversation about open data, then join in the debate on Twitter using #summerofdata.

Susanna-Assunta Sansone is an associate director at the Oxford e-Research Centre at the University of Oxford, and honorary academic editor for Scientific Data. Follow her on Twitter @biosharing.

Join the Higher Education Network for more comment, analysis and job opportunities, direct to your inbox. Follow us on Twitter @gdnhighered.