Completed on 8 Dec 2013 by Thomas Wachtler.
Login to endorse this review.
This paper describes a repository of datasets of retinal activity recorded with multielectrode arrays. The data have been collected from different laboratories. They contain spontaneous activity recorded in the context of different studies from mouse or ferret at different ages, wild type or genetically manipulated, recorded under different pharmacological manipulations or control conditions. All datasets are from published studies but have not been available publicly so far. The repository is stored on the CARMEN portal.
There is strong interest in understanding the processes underlying activity patterns in the developing retina and their relevance for visual processing. Therefore such a curated collection of retinal recordings can be expected to become highly useful, especially since the authors intend to keep extending the repository. The paper is written clearly and provides a basic overview over the data currently included in the repository. The data are accessible and provided in a simple format with sufficient documentation to read the data from the files.
I have one main concern with the paper, which is related to the authors' decision to include only a minimal set of metadata with the data. It is true that information about details of the experiments can be obtained from the original publications. However, such a paper should provide a sufficiently detailed overview to help in a first assessment of whether the data might be suitable for a given purpose at all - without having to download all the data first. At least information that is not available directly from the publications should be given, for example the actual amount of data included (for example mean firing rates in addition to duration and number of spike trains), separately for each of the experimental conditions in each dataset.
An essential piece of information that needs to be provided in any case with spike train data is the quality of spike sorting for each dataset. Some of the papers mention both multi-unit and spike-sorted single-unit activity, and there is no way of knowing what kind of spike trains we are looking at in the data files without further information. There should be a table available that shows for every dataset whether it is single or multi unit data and list the spike detection and spike sorting method, if applicable. Ideally the information should be included in the metadata, and the issue should be addressed in the paper. To avoid misunderstandings, instead of writing about "neurons'' it would be more appropriate to use the terms "units" or "recording sites'".
Fig. 3: Is there a reason why the y axis in the correlation plot is logarithmic, while in the correlation plots in the other figures a linear axis is used?
Fig. 6: The plot of burst durations could benefit from a logarithmic x-axis. It would be useful to give the number of bursts included in each distribution.
Not all Tables are referenced in the text.
Level of interest: An article of importance in its field
Quality of written English: Acceptable
Statistical review: No, the manuscript does not need to be seen by a statistician.
Declaration of competing interests: I declare that I have no competing interests.