![\"\"](\"/assets/images/figshare_bridgedb.png\")
Open Science doesn’t make publishing easier. That that’s all for the better: our research efforts are complex,\nso why should the publishing be. Sure, I am not talking about references formatting or moving the Methods\nsection to the right location, or some silly statement that all authors agree with the manuscript when you are\nthe only author.
\n\nNo, let’s talk about data. What should you publish? How, and when? And why would you do it in the first\nplace? This is not going to be a post about FAIR either, but instead about when to publish data as additional\nfiles (aka supplementary data), raw data, processed data, as a datasets, or even as a database. That’s a\nlot of types of data, and the differences matter at least for the effort you want to put in.
\n\nFirst, things have changed. We produce a massive amount more data. In the past your data, or at least the\nprocessed data, would be part of your conference talk, your journal article, or your book (chapter).\nOpen Science has changed this: data should be easier to reuse. But that results in new questions; those\nas in the previous paragraph. So, let’s add some context.
\n\nData is very broad and includes digital knowledge. Data can be raw, and the exact numbers collected (e.g.\nby a apparatus) or created by researchers. Processed data is what you get when you process the raw data.\nFor example, raw data may be a FID graph in nucleic magnetic resonance, while processed data would be a\nplot showing intensities versus chemical shifts. Published data is then a list of peaks you put in your\nresults section to support your claim of chemical identity.
\n\nA fourth type of data is metadata, and could here be the instrument on which the FID was measured, or\nthe solvent used, etc. This is where it gets complicated, because depending on the researcher who\nprocesses the data, metadata can actually be data itself. For example, when you study the chemical\nshift differences in different organic solvents.
\n\nFrom a more social level, the Open Science 101\nuses the following categories: primary data as collected/recorded by the researcher, and\n“secondary data typically refers to data that is used by someone different from who collected or generated the data”.\nThis angle of data captures the collaboration aspects of open science, but says more about\nthe processors than the data, I think.
\n\nCentral aspect of doing research is to disseminate the research. Traditionally, this has been\ndisseminating results, hoping they become facts. Increasingly, we realize that this process needs\nimprovement, particularly clearly studies, done, and communicated by the Open Science approaches.
\n\nComplementary, there is recognition&rewarding (R&R) and the wish to use various kinds of monitoring to\nassess who should be rewarded (and who should be fired), and the monitor is the implementation\nof the recognition. So, how does this work for open data? We can count every open data, but\nif thrown on a big pile, that becomes a bad monitor for use in recognition and rewarding.
\n\nOne idea is to differentiate in what data we monitor? Just raw data? Or processed data?\nHow much intellectual effort does that have to in collecting/recording the data? Should that\nbe part of the monitor and how do you even measure that? Lot’s of known unknowns here.
\n\nBut this should not inhibit us from telling the research narative. And maybe we should\njust exploring the possible narratives to allow us how it may help us monitor work done,\nhow to recognize contributions to the scientific record, and how to use all that in R&R.
\n\nI here present some example from my own research, just to start a narrative.
\n\nOver the years I have collected and recorded quite a bit of raw data. First data collected in the lab\nand later mostly recorded. Even though I have been doing Open Science since the late nineties,\nI cannot say all my data has been archived well. Even less so, I do not have a “publication list”\nof all my raw data. As an academic community, we have been focusing too much on the scholarly\narticle as the center of the research system (more on that later, because there is awesome\nresearch presented at the Dutch National Open Science Festival).
\n\nAs is defined in the European laws around GDPR,\nprocessing “includes the collection, recording, organisation, structuring, storage,\nadaptation or alteration, retrieval, consultation, use, disclosure by transmission, dissemination or otherwise making available,\nalignment or combination, restriction, erasure or destruction of [..] data”. As you can see, this is slightly\ndifferent from the first, but in light of protecting citizen, this broader definition makes sense.\nMy point here the that processing should be taken broadly. And data curation, which researchers\nroutinely do, is processing too. For any data scientist, this is easily taking up 25% of the\nfull time needed for any data analysis. One of the points of the FAIR principles is to keep\nthat number as low as possible, but not really the point here.
\n\nWhen it comes to this kind of data, I like people to have readily access to the results\nof my curation. You will find a lot of processed data like this archived. Some examples of\ndata by me or to which I contributed:
\n\nThe last one will look something like this:
\n\n
And then we have published data, which refers to data presented in a publication, like a journal\narticles. We know this as supplementary data or additional files. Several publishers, like\nBioMedCentral, submit these data automatically to a repository. For example, the\nJournal of Cheminformatics publishes all additional files under a CCZero license on Figshare.\nBut many of these support the narrative of the story, rather than the narrative of the\nresearch question. Of course, journals also have limited expectations of the format and\nmy personal impression is that these are not commonly FAIR. (Open Access is not Open Science.)
\n\nSome examples of such datasets where I do not see them as notable and do not expect them\nto be monitored. These datasets are part of the journal article, and that narrative is\nalready monitored.
\n\nAnd then we have databases provides as interactive website. This allows other researchers\nto explore the data, before the start processing the data. These typically do not have a DOI itself,\ntho data can be routinely archived as in the above WikiPathways example.
\n\nDatabases itself, as research output, are much harder to archive. And to make them citatable,\nresearch publish journal articles with a narrative that describes the database. The follwing two\nare such database papers, where the article DOI is a proxy for the database:
\n\n