HelenaLC
commented
3 weeks ago Agreed. To the best of my knowledge, this is currently missing in SpatialData (Python). The topic came up with Luca in a recent call. One option would be to stash element-wise metadata (i.e., for every image/label/point/shape/table), which could be accessed in R as a table, and used for easy filtering (e.g., give me all elements from batch A or section 2 etc.). Ideally, we allow for full flexibility here (because any dataset could, in principle, have data from many hospitals, patients, sections, technologies etc.). Minimal entries would simply be the element identifier, I guess.
The merfish.zarr in inst/extdata has zmetadata but I do not see anything that would constitute experiment metadata -- what is the source, what version of platform, etc. Can we state a minimal standard for experiment provenance and ensure that all our data artifacts can be checked for compliance?