v0.0.4-alpha Pre-release

Merged various changes from the dev branch:

Major updates

• Added process cnmf_qc for program level QC metrics as well as additional clustering metrics
• Fixed issue with scVI batch corrected counts not being batch corrected
• Migrated decoupler to python, easing python 3.8 dependency
• Added option to use scVI denoised HVGs for inferring usages, so they are batch corrected

For full changelist, see changelog

v0.0.3-alpha

Merged various changes from the dev branch:

Major changes

• Added support for scVI
• Added support for filtering biotypes using regex pattern (e.g. "protein_coding|lincRNA")
  • Filtering is applied per batch prior to merging
• Added option to completely override cNMF input files (skips all preprocessing)
• Updated way gene id conversion is configued so its much more intuitve. It now works with input and output namespaces instead of booleans.
  • manifest convert_ids > namespace (either 'ensembl' or 'gene_name')
  • config now has 'output_namespace' (either 'ensembl' or 'gene_name')
  • convert.convert_gene_ids is retired as conversion can be skipped if the namespace is matched
  • enrich.input_namespace controls ID conversion for enrich workflow (has no manifest)
• Updated maximal pathway size for GSEA & ORA from 500 to 2000

See the changelog for all details