search

drighelli / SpatialExperiment

54 stars 18 forks source link

Deprecate BumpyMatrix functionality for molecule-based data #141

Open lmweber opened 1 year ago

lmweber commented 1 year ago

We previously included some functionality using BumpyMatrix to store molecule-based spatial transcriptomics data, along with an example dataset from the seqFISH platform in the STexampleData package.

This is now obsolete, since there are now two other great packages in Bioconductor (MoleculeExperiment and SpatialFeatureExperiment) contributed by other members of the Bioconductor community, which both extend SpatialExperiment to molecule-based data in a much more sophisticated / detailed manner than our BumpyMatrix example.

So I think it makes sense to deprecate and remove the BumpyMatrix functionality from the SpatialExperiment package. This will (i) avoid confusing users and (ii) reduce dependencies.

Any thoughts? I'll assign this to me for now.

drighelli commented 1 year ago

Hi Lukas (@lmweber),

I'm glad you raised this point because I'm also interested in this aspect.

I saw the MoleculeExperiment and the SpatialFeatureExperiment packages, but still, I think it's too soon to deprecate our implementation in favour of other external packages.

More in detail, I'm not sure that those implementations are better from a computational point of view in handling molecule-based data. We don't have any benchmarking where the performances are compared yet.

At the moment, I'm working with multiple molecule-based data, focusing on some platforms and I'm trying to do some implementations starting from the actual implementation of SpatialExperiment, MoleculeExperiment and SpatialFeatureExperiment and I'd like to test their performances.

But until we don't have any comparison in terms of memory usage, computational elapsed time, etc., I would not deprecate any functionality in our or other "spatial" package.

It seems that we are at the beginning of the spreading of these technologies and I think so much more approaches will come out soon.

lmweber commented 12 months ago

Hi @drighelli , ok thanks, that sounds good!

If you are trying out and comparing some of these implementations with some real datasets then this will hopefully also help us get a better idea of how these implementations compare.

For now I believe it is mainly these three implementations available in Bioconductor for storing molecule-based data: SpatialExperiment, MoleculeExperiment, and SpatialFeatureExperiment. It will also be useful to keep a look out for any additional new packages that build on each of these.