 I'm Ian Trinaskiewicz from Nature Publishing Group and Pagode McMillan, which is now part of Springer Nature. The editorial policy of the Nature Journals, which I have on screen here, is for at a bare minimum data to be made available to editors, reviewers and readers without undue qualifications, so data on request. However, the strong preference of the Nature Journals is for data to be available in public repositories wherever possible, and for those repositories to be specific to particular types of discipline or subjects, and of course there are mandates in place for certain scientific communities where data must be made public, and we have a list of the types of data and the supporting repositories where there are mandates in place. To support our authors in getting their data into a public repository for their discipline wherever possible, we recommend that they consult the list of recommended data repositories that is managed by the Scientific Data Journal team, and so this policy lists and links to a list of approved repositories on the Scientific Data website where we encourage authors to find an appropriate home for their data. Where a subject specific repository does not exist, then general repositories such as FigShare and Drive are recommended as alternatives. So given our multidisciplinary coverage of different areas of science and the number of different journals, the types of repositories and the types of data underlying publications are quite varied. Historically, one would be more likely to find references to data sets in the text of articles in the types of articles and the types of research where there are mandates in place such as these that I'm highlighting here. As well as data and repositories, I think it's important not to ignore that supplementary information files can still currently hold data that support publications, but in terms of type of data, it's probably easier to think about the types of file that supplementary information tend to be. So commonly types of file are spreadsheets, video files, and PDFs. Of course, supplementary information files can hold text as well, such as supplementary information, supplementary methods. In terms of how supporting data sets are referenced in nature articles, we have a mixed methodology I would say at the moment. I'm going to show a couple of examples. For the most part, data that have particular types of a session code are referenced in a dedicated section of the text, and we have robust linking rules in place where particular types of a session number or data set identifier are recognized, and then hyperlinks to those are created as part of the production process. So looking at this example from Nature Communications, we have an accession code section of the article, and in this case, we have three different types of referenced sessions. So from Gembank, Geo, and the Sequence Read Archive, with the hyperlinks all in place, and this is above the reference list. We also have another example here with the session codes, and in this case, the reference of sessions are from the Protein Data Bank. We also have another approach in place, which is from Scientific Data, which is a data journal, and we actually have formal data citations in all articles and scientific data. So these follow immediately after the reference list, and so this is in compliance with the Joint Declaration on Data Citation Principles, and this system enables us to support a much wider array of data set identifiers, including those such as this one from Dryad, where DOIs are associated with data sets. The format here is Author, Repository Name, Identifier, Withalink, and then the Year. So we have a number of some mixed methodologies at the moment, but I think we need to work towards more standardized solution and a way for articles to support the formal citation of data sets in reference lists. So I would say that the data references and citations are certainly reasonably human readable at the moment, but the machine readability could be further improved. So just to give you one other example, Gigascience, which is published by Bioman Central, which is now part of Springer Nature, also have formal citations to data in the Gigascience repository in a dedicated section of the article, such as this example here. So in terms of what we can do better, I mentioned the Nature Publishing Group and in fact Bioman Central have endorsed the joint declaration on data citation principles, and longer term we would want to find a robust solution that would ensure the quality and usefulness of citations to data across all these different kinds of repositories, but they're useful both for human readers so that they can understand where the data are and find the links, but also for machine readers when we're looking to harvest and collect up data article citations and associations. The two main dependencies for implementing a better system are the editorial policy and the processes for encouraging or requiring data to be cited across all of these different disciplines and the technological enhancements to make this happen in the ideal way. On technology, we recognize that changes to XML structure and article presentation are needed, specifically implementation of the journal article, taking sweet JAPS 2.0, which for a large publisher with lots of different journals and imprints, that kind of work does take time and is a significant undertaking. On the process and policy side, I think there are particular challenges which perhaps could warrant further discussion in the publishing and research community as the best ways to handle them. In particular, where we have articles that have a very large number of data citations such as articles that prevent a large number of accession codes, how is the best way to handle those in article reference list, particularly when a number of journals do still face real limitations in terms of being printed as well as being online. I also think that we may have to engage authors with discussions about citing works. There may be academic preferences and cultures about the inclusion of references, particularly things like the very large number of accession codes in article reference list, which we can't ignore. Another more complex issue potentially which I'm not sure any publishers come up with the ideal solution for yet is how we differentiate between data that are generated by a study from data which are analyzed by a study and separating those also from data which are merely referenced by a study. There are three different situations there which the ideal data citation system would aim to capture and address so that they can make those kind of article data links and then your answers clear for both the machine readers of the literature and for human readers who are often just trying to find the links between articles and data. Thank you.