At the University of Leeds we've been giving this some thought. There are a lot of scenarios to deal with, so the right solution as nearly always depends. Let me outline one of our cases briefly and give you my reckoning which is not necessarily that of my institution or my other colleagues working on this.
ForestPlots.net have amongst other things soil sample physical stores. Some physical samples have not been catalogues yet, but most have. Most physical samples have been analysed in some way in a laboratory. This generates digital metadata for them and a digital profile for the physical samples is created. These are stored collectively in a form of database.
Assigning DOIs for individual samples (bags of dried soil) does not scale in the Forestplots.net example. Some of the samples we deal with are collections from specific surveys. These surveys may be for area that are resurveyed or from a transect that is not planned to be resurveyed necessarily. Some 'plots' are monitored in an on-going long terms fashion. In general for ForestPlots.net it makes sense to assign DOIs for each survey and for each plot and for collections of plots. At the moment, a single DOI has not been assigned for all the ForestPlots.net data, but as a pilot ForestPlots.net has its own DOI prefix.
The appropriate granularity of DOI assignment all depends.
The issue of having large author lists on journal articles is similar to the problem of having lengthy reference lists. We have debated the creation of DOIs to represent all the ForestPlots.net data that is used for the results reported in a particular journal article. If the article uses more than just these data, then it can refer to other data in the same way. One of our team was keen to avoid the creation of a single DOIs unique to each journal article which refers to all the data used for that. I think though that in some cases this might be the way forward to mitigate the problems caused by having too long a list of references.
We debated using the suffix of a DOI to represent a hierarchical structure of the data, but I think for the time being, the DOIs are simply numerical and the detail of what the DOI refers to is metadata to be found via the DOI landing page.
I hope that is a useful and accurate summary and helps moves the discussion forward. Apologies in advance if I've got anything wrong.
From: Research Data Management discussion list [mailto:[log in to unmask]] On Behalf Of M. Casula
Sent: 03 April 2014 09:23
To: [log in to unmask]
Subject: assigning DOIs to tissue samples
Some months ago I posted a question about assigning DOIs to samples and got some useful feedback. Having participated at the 'Internet of samples' session at the recent RDA plenary, it is clear that this topic is of general interest with representatives from the earth, agricultural sciences, life sciences and engineering sciences participating at this session. Having now set up a prototype for the DOI enabled site for tissue samples, I have now run in another conundrum on which I would be grateful for advice. Having discussed the assignment of DOIs to tissue samples with colleagues, they have raised a very valid issue in the circumstance of the results form several hundred tissues samples being reported in the same article. If current recommendations are followed such as those of DataCite, then the citation format for datasets is very similar for that of conventional written publications, in which case the reference section would extend to several pages. While I am aware that DOIs can be assigned to data collections and that supplements to articles are also becoming common, neither of these solutions are particularly appropriate. In the first case it would mean having to define a collection for every article which is not practicable and in the second instance the data are becoming dissociated from the written article which is basically counter to the purpose of data citation. As this is surely a potential issue in any discipline, if anyone is aware of an alternative solution I would be grateful for further information.
AMC Disclaimer : http://www.amc.nl/disclaimer