I have an ExperimentData package that I am working to convert into an ExperimentHub package. The package is called sceptredata and is available on Github.

One of the datasets contained within sceptredata is a single-cell dataset stored in 10x Cell Ranger feature barcode format. The data are located in the inst/extdata/highmoi_example directory of the package. The directory structure for the example data is as follows:

- highmoi_example
     - gem_group_1
         - barcodes.tsv.gz
         - features.tsv.gz
         - matrix.mtx.gz
     - gem_group_2
         - barcodes.tsv.gz
         - features.tsv.gz
         - matrix.mtx.gz

It is important that the data be stored using these file names and using this directory structure. Is it possible for me to preserve this directory structure upon converting my package into an ExperimentHub package? Should I zip the directory highmoi_example so that it can be treated as a single file (and then somehow programmatically unzip highmoi_example upon download of the dataset)?

In summary, is there a protocol for handling data that have some directory structure in the context of ExperimentHub? Thanks in advance for anyone who is able to chip in.

Hi, this type of question is probably better suited for the Bioconductor devel mailing list. I think what you may want to do is to implement a wrapper function in your ExperimentData package that downloads the ExperimentHub resources and copies and organizes the files in the way you need; see for instance the vignettes in the TENxPBMCData and the gDNAinRNAseqData experimental data packages, where I developed the latter myself.

Yes agreed. Either a wrapper around organizing the way you need. Or another possible option would be to maybe tar/zip the files all together in one download and then once downloaded local unzip/un tar. That might result in a large download that might require some guidance on systems for adjusting a timeout limit depending on download speed but could be another option too.