Portal de Periódicos da CAPES

Opening up: open access publishing, data sharing, and how they can influence your neuroscience career

2016; Wiley; Volume: 43; Issue: 11 Linguagem: Inglês

10.1111/ejn.13234

1460-9568

Tara L. Spires‐Jones, Panayiota Poirazi, Matthew S. Grubb,

Genetics, Bioinformatics, and Biomedical Research

No matter how often it might feel like it, science doesn't happen in a bubble – the things we discover aren't worth anything unless they influence other people. Luckily, this ability to reach others with our science is being made ever easier in the digital era. We have already moved far from the traditional model in which scientific findings were only ever presented in subscription-access, print-only journals. Nowadays there are huge opportunities to widen access, not only to complete scientific articles, but also to their underlying raw data, and there is a growing push from funders and other stakeholders to promote such openness. Here, we aim to outline these increasing demands for open access (OA) publishing and data sharing, to describe the routes available for their implementation, and to weigh up the costs and the benefits associated with such scientific openness, especially for early-career researchers. We don't claim to be exhaustive in our coverage of this large subject, and refer interested readers to the many excellent and comprehensive reviews and opinion pieces written by others and cited in our reference list. But we do aim to bring a neuroscientific, and particularly a young neuroscientists' perspective to the issue, and in this goal we are helped throughout by the responses of our own FENS-Kavli Scholars to a simple questionnaire we gave them in the autumn of 2015. In the spirit of openness – of course – this questionnaire and the grouped responses to it are freely available as supplementary data to this article. We're also continuing to collect responses from an online version, found here (https://fkne.typeform.com/to/Jx4NAq) and open until a year post-publication, in which we would greatly appreciate your participation. With enough additional data, we hope to be in an even stronger position to lobby and advise relevant parties on some of the crucial points we raise below. There is a prevalent and ever-growing view that publicly-funded research should be accessible to everyone. This is reflected in increasing demands from funding bodies and other relevant parties that neuroscientists be open with both their publications and their data. Funders, in particular, are increasingly requiring scientists to publish OA articles, and to deposit raw research data in public archives as a condition of funding. This is evident by growing efforts in both Europe and the US whereby public funding institutions have adopted OA requirements in their funding schemes (http://tinyurl.com/j62khsp). These requirements are sometimes – but not always – enforced strictly. For example, the NIH in the U.S. and the Wellcome Trust in the U.K. can take extreme measures such as freezing of funds if publications from funded projects are not made OA. For the European Research Council (ERC) and the Research Councils UK (RCUK) on the other hand, while OA of publications is a requirement, no enforcement plan is clearly evident yet. In Europe, Open Science is the new initiative adopted by the European Commission (http://tinyurl.com/q7bqkxf). Commissioner Moedas and Commissioner Oettinger in a joint blog post on June 22, 2015 stated that 'Open Science describes the on-going transitions in the way research is performed, researchers collaborate, knowledge is shared, and science is organised. It represents a systemic change in the modus operandi of science and research. It affects the whole research cycle and its stakeholders, enhances science by facilitating more transparency, openness, networking, collaboration, and refocuses science from a 'publish or perish' perspective to a knowledge-sharing perspective.' Towards this goal, the Commission has already initiated several science policy actions, with two large OA pilot initiatives currently operating at the EC level: (i) the FP7 post-grant OA publishing funds pilot (http://tinyurl.com/zzq9lur), where researchers and/or organizations can request funds to cover the OA publication fees for outcomes of a completed FP7 grant, and (ii) the Open Research Data Pilot (http://tinyurl.com/jn4deqz), which 'aims to make the research data generated by selected Horizon 2020 projects accessible with as few restrictions as possible, while at the same time protecting sensitive data from inappropriate access.' Similar initiatives at the National level include the recent announcement (http://tinyurl.com/q3yzdtv) of the Dutch National Research Funder NWO that makes OA mandatory. In essence, NWO demands that all publications emerging from a 'call for proposals' published by NWO after December 1st 2015 must be immediately accessible to everybody from the moment of publication. According to the announcement, NWO is the first national research council worldwide to take such a step and, importantly, it is supporting this transition to OA financially. Within our own group of 20 young neuroscientists working in Europe, 55% (n = 11) have received funding from institutions that insist on OA publications. However, the funds dedicated to such publications were only available to 25% (n = 5) of the scholars. Although the sample size of our own little survey is admittedly very small, these results indicate that while OA is largely promoted via funding agencies, it is not yet fully financially supported. Overall, the current demands of funding bodies for OA of research outcomes are highly variable. This tends to be far stricter for OA to publications rather than for data sharing at the moment, largely because data sharing requires the development of appropriate digital infrastructures and management protocols which are currently underway. However, it is becoming more and more apparent that OA and data sharing, or 'Open Science' is a rapidly approaching reality and that everyone involved should be prepared for the inevitable. In addition to funding bodies and organizations, pressure for OA comes from other sources. The Max Planck Society has been actively advocating OA since 2003 (http://tinyurl.com/htjoz37), with the publication of the 'Berlin Declaration on Open Access to Knowledge in the Sciences and Humanities.' Ever since, it has been holding annual conferences to increase awareness and propose measures towards OA. In Canada, McGill University's Montreal Neurological Institute (MNI) has recently embraced openness on an unprecedented institutional scale – on a voluntary basis, its researchers are making all results and data freely available at the time of publication, and are foregoing patent applications. What's more, in the hope that this approach will go viral, they're requiring their collaborators to sign up to the same principles (http://tinyurl.com/z7lu4do). There are also a number of collective organisations promoting scientific openness. These include Voice of Researchers (VoR; http://voice.euraxess.org), a network formed in 2012 that takes an active role in shaping the European Research Area. The League of European Research Universities (LERU; http://www.leru.org), a prominent advocate for the promotion of basic research at European universities, is also a strong supporter of OA. In October 2015, LERU issued a statement entitled 'Christmas is over. Research Funding should go to research, not to publishers!' (http://tinyurl.com/jhlqqkk), proposing a new business model in favour of OA. Specifically, 'LERU wants universities, which pay for subscriptions, to be able to use their current spending level to 'offset' subscriptions against payment for article processing charges (APCs) for journal articles in hybrid journals. As part of any agreement, publishers should permit all papers published by university researchers taking up the deal to be made open access for no extra charge.' And ERCIM, the European Research Consortium for Informatics and Mathematics (http://www.ercim.eu), has recently published a report entitled 'BOM@ERCIM — Towards an open access policy for ERCIM'. This document provides a basis for better communication between research organisations about dissemination of research results and OA. It also gives a strong set of recommendations that could be implemented step-by-step by all ERCIM members and other interested parties. Importantly, in addition to public institutions, private enterprises have also joined the quest for OA and data sharing. The Allen Institute in the U.S. has adopted an openness attitude from the beginning (e.g. via free access to its brain atlases) and has recently joined the effort to generate the necessary protocols and digital infrastructure that would allow large scale data sharing. The neuroscience community in particular is a major driving force in this effort, primarily due to the funding invested and the massive amounts of data generated as part of the two large neuroscience initiatives: the Brain Initiative in the U.S. and the Human Brain Project in Europe. With respect to data sharing, several journals and commercial repositories are also pushing open data forward. For example PLoS journals now require open data associated with their publications as detailed in their blog from Feb 2014 (http://tinyurl.com/hpmp6xm). They 'strongly encourage deposition in subject area repositories…where those exist, and in unstructured repositories…where there is no appropriate subject-domain repository.' In line with these positions, our own group also managed to agree on something: 100% of the FENS-Kavli Network of Excellence (FKNE) Scholars (n = 20) believe that OA to both data and publications is good for neuroscience. With everyone pulling in the same direction, and with openness becoming an increasingly common requirement for scientific endeavour, the natural next question is 'how do I do it?' OA publication is widely available, particularly since the government and funder mandates mentioned above have come into effect. However, the world of OA publishing can still be quite confusing, with different grades of openness available. To help clarify things a little, we spoke to Phill Jones, head of publisher Outreach Digital Science and a 'chef' on the Scholarly Kitchen blog (http://scholarlykitchen.sspnet.org). He told us that, traditionally, authors signed over the copyright of their publication to the journal, which charged subscription fees to access their publications. This is still the case in many journals, but because of OA requirements most of these journals now allow 'self-archiving', which means that the final accepted peer-reviewed version of the article, but not the publisher's nifty pdf version, can be uploaded to a repository such as PubMed Central or institutional websites after an embargo period. The embargo period is typically 6–12 months, meaning that only people with subscriptions or those willing to pay the fee to buy the individual article can read the paper during this initial period after publication. This self-archiving method is often referred to as 'green' OA publishing (Table 1). A level up from green is 'gold' OA which allows immediate access to the published final version on the journal website. Often gold OA is paid for by the author through high publication fees, in the order of several thousand euros. Even within gold OA, there are grades of openness defined by the copyright license. The most open publishing copyright commonly used by journals is the Creative Commons Attribution license (cc-by), which permits anyone to read, distribute, or reuse the article as long as the original source is properly cited. Still OA but slightly less so are articles published under non-commercial (NC) and non-derivative (ND) cc-by copyright which do allow free access to the final published article but do not permit commercial use of the article (NC) or derivative re-use like text mining or data mining (ND) (for more information about the confusing area of licences, see http://creativecommons.org/licenses). Most purely OA journals, such as eLife, Frontiers and the PLoS journals publish with the cc-by license. To make matters more confusing, many journals (such as Nature, PNAS, and EJN) now have a hybrid model where most of their articles are not OA or can be green OA after the embargo period, but authors have the option of paying extra to make their papers immediately accessible with gold OA. This can be particularly infuriating to universities and other institutions because they are effectively paying twice for access to articles, once through subscription fees and again by paying for gold OA. So now we know a bit about what OA is, but is it important for us as early-mid career neuroscientists? Within the admittedly small sample size of the FKNE, 100% (n = 20) of the scholars agreed that OA publishing is good for neuroscience. However, only 52% (n = 183/354) of our collective PI-authored papers are currently freely available online. When asked 'How important are open access options in your choice of journal for your lab's primary research publications?' only one scholar responded with the choice 'it means everything to me'. The most common choice (40%; n = 8) was 'it enters my thoughts briefly and then I go back to checking impact factors', while a close second (35%; n = 7) responded 'it definitely plays a role in the decision-making process' and 20% (n = 4) thought 'it really doesn't matter'. For neuroscientists, the most important thing to keep in mind about OA publishing is whether you are complying with your funder's requirements. So before deciding which journal to publish in, it's important to make sure that if your funder requires gold OA, that is an option in your journal of choice. Some funders, such as the Wellcome Trust and RCUK in the UK, provide funds directly to universities as 'block grants' to pay the OA fees (see below). Sometimes, your institutional librarians know about these opportunities and can help you pay for OA fees. Many funders require green OA, in which case it is your responsibility to make sure the final accepted version of the article (not the publisher's pdf) is shared online. This can be done through public repositories like PubMed Central, ResearchGate or OpenAIRE (Table 2), or often through institutional repositories. Again librarians can often help with this process and inform you whether your institution has any particular rules. Beyond sharing the final version of published articles, sharing raw data is another important aspect of open science. Data sharing has been particularly successful in some areas with standardized data outputs such as genomics (Choudhury et al., 2014). For instance, the International Nucleotide Sequence Database Collaboration (http://www.insdc.org) combines worldwide genetic sequence data including the European Nucleotide Archive, the US NIH GenBank, and the DNA DataBank of Japan. In neuroscience, the Psychiatric Genomics Consortium (http://www.med.unc.edu/pgc/) has freely available genetic data from over 900 000 people with psychiatric disorders and controls. The Human Connectome Project (HCP; http://www.humanconnectomeproject.org) is another successful neuroscience data sharing initiative that collects and shares MR imaging, cognitive, and demographic data on 1200 healthy volunteers to define variation in brain wiring (Van Essen, D.C., Smith, S.M., Barch, D.M., Behrens, T.E.J., Yacoub, E., Ugurbil, K., & WU-Minn HCP Consortium, 2013). PubMed shows over 1000 hits for this project indicating its utility to the field. There are also many other general or specialised repositories for brain imaging data, where sharing appears to be particularly well resourced (Eickhoff et al., 2016; Table 2). In addition, the Collaborative Research in Computational Neuroscience (CRCNS; https://crcns.org) data sharing repository is a joint effort of the National Science Foundation (NSF), the National Institutes of Health (NIH), the German Federal Ministry of Education and Research (Bundesministerium für Bildung und Forschung; BMBF), the French National Research Agency (Agence Nationale de la Recherche; ANR) and the United States-Israel Binational Science Foundation (BSF) that supports the sharing of experimental data from various brain regions, analysis tools and simulations. Another incredibly useful neuroscience data sharing initiative comes from the Allen Institute for Brain Science, which freely provides multiple large datasets in the form of brain atlases highlighting development, connectivity, and cell types in several species (http://brain-map.org). According to Terri Gilbert at the Allen Institute, they have more than 45 000 visits to their brain maps every month, around 150 citations per year of the mouse atlases, and 25 citations per year each of the connectivity and human atlases. Digital reconstructions of neurons are another example of successful data sharing in neuroscience. Digitized neuron traces are shared on NeuroMorpho.org (Parekh & Ascoli, 2013; Ascoli, 2015), and the Allen Institute is working on the BigNeuron project to provide a resource for automated neuronal reconstructions from imaging data (Peng et al., 2015). With respect to sharing of computational models of neurons and circuits, ModelDB is the largest open repository at the moment (https://senselab.med.yale.edu/modeldb). For electrophysiological recordings, the Carmen portal (http://www.carmen.org.uk) provides an open repository as well as analysis tools, and the Neurodata Without Borders (NWB; http://nwb.org; Teeters et al., 2015) project continues to develop a common data format for ease of sharing. In addition to these large initiatives sharing standardized data sets, there are options for sharing any type of data. There are unstructured repositories such as Dryad, FigShare and Zenodo, and many institutions provide their own repositories for their researchers' data. Data sharing should also be greatly facilitated by the EU's OpenAIRE initiative (https://www.openaire.eu). OpenAIRE is a network of OA repositories, archives and journals that support OA policies. It goes beyond the traditional publications aggregator by interconnecting entities related to scholarly communication (publications, research data, funding, people, organizations, data sources) allowing users to navigate alongside a rich information space graph and provides a wide range of services, from deposition to statistics. At the time of writing, OpenAIRE provided access to 13 333 818 publications and 16 735 datasets from 6081 data sources. These involved 98 410 projects and 18 351 organizations! These repositories, amongst others, are summarised in Table 2. This can be used as a first stop for choosing a sharing route, but is not by any means exhaustive. For the full picture of your data sharing options we recommend the Neuroscience Information Framework (http://www.neuinfo.org), a comprehensive 'database of neuroscience databases' set up by the NIH's Blueprint for Neuroscience Research. However, even with so many opportunities available to share and exploit neuroscience data, their use is pretty low amongst our Network: 35% of us (n = 7) have not shared any data openly yet, and only 45% of us (n = 9) have ever used shared data to benefit our own research. However, there are examples within our Network of sharing several types of data, including sequences, behaviour, electrophysiological data, morphological data, and computational models. One key question remains concerning the sharing of neuroscience data: what exactly should be shared? In other words, what is the 'minimum data unit' that is useful for sharing with other scientists? This clearly has huge implications, on the one hand for the work that our fellow scientists are able to perform with the data we have generated, and on the other for the amount of time and effort required to get our data in sharable form (see below). But there's little agreement as to the most useful balance to be found here. Amongst the FKNE Scholars, 50% (n = 10) thought that defining the minimum data unit was 'hard to say, and depends on the type of data', while the remaining votes were split evenly between 'all raw data that contributed to a single publication' (20%; n = 4) and 'all raw data represented in the figures of a single publication' (25%; n = 5). Defining the standard in the field will be a crucial step to advance data sharing in neuroscience in the near future. Being open with articles and data is therefore becoming increasingly demanded of neuroscientists, and the options for fulfilling these demands are growing and growing. But what are the benefits, and the potential drawbacks, of openness in modern neuroscience? How can it influence your career, especially if you're just establishing your own lab? Many people think the answer to both these questions is 'not much'. Asking our Network revealed very little expectation that OA publications or data sharing would impact on the careers of young neuroscientists. Only 35 % of us (n = 7) agreed that 'publishing open access articles is important for the career development of early-stage neuroscientists'. And while asking for 'good reasons to share your data openly' revealed some very noble motives amongst our group – the top answers were 'It's good for science' (n = 18) and 'It's the right thing to do' (n = 12) – only six of us thought that 'It benefits my career'. However, there are some clear and demonstrable benefits to being open with your science. Perhaps the negative perception of the career impact of openness is because so few of these benefits are 'direct' – where publishing OA articles or sharing data are rewarded in their own right. But direct benefits do exist, even if they are only permissive. For example, funding bodies such as the NIH and the Wellcome Trust now insist that their researchers publish their findings in OA format (see above), and can prevent future applications from those who do not comply. In other words, publishing OA articles brings the benefit of being eligible for future grant applications. While these eligibility criteria are not currently so strict for other funding bodies, many, including the ERC and RCUK, are known to be reviewing policy in this area. For data sharing, the guidelines and requirements are currently nowhere near as strict as those for OA publishing (see above). However, many grant applications now require researchers to list their plans for data sharing, and in today's ultra-competitive environment this is, at the very least, an area where a lack of detail or poor motives could pull a project just under the 'fundable' bar. Moreover, in the EC's 2014 public consultation on Science 2.0: Science in Transition (http://tinyurl.com/hh7fegv) 85% of respondents (48% 'totally agree', 37% 'partially agree') believed open science activities 'should be taken into account for researchers' career progression'. Researchers who can demonstrate not only high productivity of their own labs but also the facilitative influence of their data on other labs are well positioned to gain a decisive advantage over similarly productive competitors who choose not to share their data with peers. (Ascoli, 2015) Despite some very real potential costs that should always be taken into consideration, it is clear that there are manifest and multiple benefits to being scientifically open, both with your publications and with your raw data. None of these benefits apply exclusively to young researchers, but today's changing landscape will certainly have a greater impact on those with the longest careers ahead of them. Openness is undoubtedly the future, so be part of it now – it'll help you, and it'll help us all to produce some fantastic neuroscience! The FENS-Kavli Network of Excellence is supported by FENS, the Kavli Foundation, Alzheimer's Research UK, the European Molecular Biology Organization (EMBO) and Roche. PP acknowledges funding from the European Research Council (StG 311 435 dEMORY). We would like to thank the FKNE Scholars for their questionnaire input, and the following people for their thoughts in preparing the manuscript: Jean-Claude Burgelman (EC), Michele Garfinkel (EMBO), Terri Gilbert (Allen Institute), Mark Hahnel (figshare), Tim Hubbard (King's College London), Phill Jones (Digital Sciences), and Alex Saxon (RCUK). The authors are Scholars of the FENS-Kavli Network of Excellence, a network of young neuroscientists, with the goal of fostering exchange – scientific or about science policy – between excellent junior/mid-career neuroscientists who are either currently working in Europe or received their academic training in Europe. Tara Spires-Jones is a reader and Chancellor's Fellow leading a research team at the University of Edinburgh. She studies the brain changes underlying dementia with a focus on synaptic degeneration. Panayiota Poirazi is a Research Director at the Institute of Molecular Biology and Biotechnology (IMBB), Foundation for Research and Technology-Hellas (FORTH). She used computational modelling techniques to study neuronal and dendritic computations and their role in learning and memory functions. Matt Grubb is a Lecturer in the Department of Developmental Neurobiology at King's College London. His group investigates the role of electrical activity in sculpting brain maturation, with a particular focus on circuits in the olfactory bulb. Data S2. Questionnaire responses raw data. [Correction added on 24 August 2016, after first online publication in April 2016: Two supporting information were omitted on first publication and have now been added in this current online version.] Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.