KJ optionally use h5seurat instead of rds

[kathrinjansen]

• add optional usage of 'begin.h5seurat' as input instead of 'begin.rds', for this the package SeuratDisk is used

• new option in yml for this: input_format = rds | h5seurat , this option is used for multiple tasks throughout the pipeline_seurat.py

• for option 'h5seurat': instead of writing out flat files for e.g. scaled data or metadata, the begin.h5seurat is converted to an anndata object (begin.h5ad) and this is used for other tasks (new R script for this)

• fixes to code for diffusion map task (was running only once per components not for each cluster resolution?)

• test done for following combinations: from h5seurat + loom velocity, from rds + loom velocity, from raw mm format using h5seurat

[snsansom]

Hi Kathrin,

A few initial comments:

Can we please move the code for loading and saving of Seurat objects into functions in the tenxutils R library.

Please only require(SeuratDisk) if it is needed (i.e. within the functions).

Thanks,
Steve

[kathrinjansen]

There are now two new functions in tenxutils named loadSeurat and saveSeurat which load from rds or h5seurat depending on file ending or option 'format'. Package requirement removed from all other scripts.

Some more fixes to issues with s@misc in h5seurat: it's saved as a list if h5seurat is created by pipeline or if conversion happens, the gene_id to gene_name mapping is not in s@misc but in @assays$RNA@meta.features instead .

[snsansom]

Have rebased on master but the branch is not running on the test datasets (geneset analysis fails on mouse datasets, export anndata fails with input format h5seurat).