0.7.1

openpipelines 0.7.1

NEW FUNCTIONALITY

• integrate/scvi: use nvcr.io/nvidia/pytorch:22.09-py3 as base container to enable GPU acceleration.

• integrate/scvi: add --model_output to save model.

• workflows/ingestion/cellranger_mapping: Added output_type to output the filtered Cell Ranger data as h5mu, not the converted raw 10xh5 output.

• Several components: added --output_compression component to set the compression of output .h5mu files.

• workflows/full_pipeline and workflows/integration: Added leiden_resolution argument to control the coarseness of the clustering.

• Added --rna_theta and --rna_harmony_theta to full and integration pipeline respectively in order to tune the diversity clustering penalty parameter for harmony integration.

BUG FIXES

• mapping/cellranger_multi: Fix an issue where using a directory as value for --input would cause AttributeError.

• workflows/integration: init_pos is no longer set to the integration layer (e.g. X_pca_integrated).

• dimred/pca: fix variance slot containing a second copy of the variance ratio matrix and not the variances.

MINOR CHANGES

• integration and full workflows: do not run harmony integration when obs_covariates is not provided.

• Add highmem label to dimred/pca component.

• Remove disabled convert/from_csv_to_h5mu component.

• Update to Viash 0.7.1.

• Several components: update to scanpy 1.9.2

• process_10xh5/filter_10xh5: speed up build by using eddelbuettel/r2u:22.04 base container.

MAJOR CHANGES

• dataflow/concat: Renamed --compression to --output_compression.