Identify the input formats for single-cell data we want to support in Gemma

[arteymix]

It looks like there's three data formats commonly used out there:

• SeuratDisk, and HDF5-based storage for Seurat
• AnnData on-disk storage, another HDF5-based storage
• MEX a 10X format with the .mtx extension, does not include samples/factors, so an additional user-supplied mapping would be necessary
• Loom which is another HDF5-based format

[arteymix]

I'm currently implementing the import for AnnData HDF5. It will be built on top of HDF5 JNI API and reusable for Seurat.

MTX is relatively easy to do since it's just a tabular format that we can parse with Apache Commons CSV.

[arteymix]

Another format to consider are aggregated MEX which are basically multiple samples combined in a single matrix.

https://www.10xgenomics.com/support/software/cell-ranger/latest/analysis/running-pipelines/cr-3p-aggr

[arteymix]

Sample associations for MEX formats can generally be determined by the naming scheme of the submitted files which get prefixed by GSM IDs.

I haven't looked yet at aggregated ones, but I suspect we might not have that benefit.

Those nitty gritty details should be dealt with on the GEO loader.

[arteymix]

This is done. We'll support MEX, AnnData and Seurat Disk.