Changelog • scdrake

scdrake 1.6.0

scdrake now allows processing of spatial transcriptomics data from spot-based technologies (Visium).
- See vignette("scdrake_spatial").
Added annotation using user-defined marker genes.
Updated stage_input_qc and stage_norm_clustering vignettes.

Major refactoring:
- Divided some sections into smaller subplans (e.g. dimred plots or clustering). This also applies to child RMarkdown documents.
- Minor config polishing.
Changes in clustering:
- All clustering types can be now enabled or disabled.
- Added Leiden algorithm for graph-based clustering.
- Added multiple resolutions for Leiden and Louvain algorithms.
- Polished clustering names in colData().
- Added clustree to visualize cluster abundances under different ks or resolutions.
Added Harmony integration.
Allow to load data in the 01_input_qc stage also as a SCE object from a drake cache.
Subsetting of imported SCE object in the 01_input_qc stage.
Major refactoring and update of documentation (vignettes).

Fixed line ending option in the autogenerated RStudio file for new projects.
"rds" is now the default storage format for drake cache.
Added more drake-specific config parameters to config/pipeline.yaml.
- See DRAKE_MEMORY_STRATEGY which can help on machines with low memory.
Added support of arm64 architecture to download_yq().
Added a Dockerfile for arm64 platform.
Updated scdrake_docker vignette.

scdrake has been ported to a Docker image and this is now the recommended usage.
- See vignette("scdrake_docker").
A command line interface was implemented. It wraps several main scdrake functions.
- See vignette("scdrake_cli").
run_single_sample() and run_integration() are now soft-deprecated.
Several bugs have been fixed.
Text/visual/user-experience refinements.
Dev changes:
- dev/run_check.R and dev/run_tests.R (former run_pipeline_tests.R) now uses R-based CLI and are run through Rscript.
- deps.Dockerfile will install scdrake with all dependencies from DESCRIPTION.

New feature: “inject” custom cell data. See the ADDITIONAL_CELL_DATA_FILE parameter in 02_norm_clustering stage vignette.
Datasets for integration can be now imported also from Rds files. The structure of the INTEGRATION_SOURCES parameters has changed - please, review your configs.
Updated documentation.
Fixed some rare bugs.

Added support for a custom plan located (by default) in plan_custom.R script in project directory.
The GitHub version of SingleR was removed from DESCRIPTION as the version for the current Bioconductor version is available from the official Bioc repository. The GitHub version is compatible with Bioc 3.16 and caused installation error on Bioc 3.15
Fixed “object cfg not found”.
Documentation has been improved:
- Clearer instructions in the Get Started vignette. The part about the integration pipeline was moved to a separate vignette.
- New vignettes about extension of the pipeline and FAQ & Howto.
- Clearer installation instructions, including commands needed to install shared libraries on different OSs.
- Some parts in other vignettes have been rewritten, spellchecked and improved cosmetically.
Cell or gene filtering can be now disabled by a single parameter (ENABLE_CELL_FILTERING, ENABLE_GENE_FILTERING).
Require clustermq version >= 0.8.8, but show a warning if the version is greater.
Checks for SC3 version if it is not installed from github.com/gorgitko/SC3, but also allow version from Bioconductor.
Show used functions with links in stage reports.
New pipeline config variable DRAKE_REBUILD allowing to force rebuild targets.
Fixed various corner case bugs.
Small enhancements in some functions.
Some vignettes have been extended or updated.

Fixed a bug in the sce_final_norm_clustering target where already present columns in colData() were not updated by new ones in the cell_data target. This could happen when SCE object was imported in the 01_input_qc stage and already contained cluster assignments in colData() from the previous scdrake run.

Fixed a typo in 01_input_qc stage report, added a line for emptyDroplets lower bound in barcode rank plot.
save_pdf() is now able to treat invalid plots throwing an error on save.
- A new helper function create_dummy_plot().

Dimred plots of clusterings and other variables are now saved as PDF and PNG files.
- In 02_norm_clustering and 02_int_clustering stage reports the PNG files are displayed and act as links to associated multipage PDF files.
- New parameters for the stages above: NORM_CLUSTERING_DIMRED_PLOTS_OUT_DIR and INT_CLUSTERING_DIMRED_PLOTS_OUT_DIR, respectively.

Added cell type annotation via SingleR (https://github.com/LTLA/SingleR).
- See the new config parameters here and targets here.
Code and documentation refactoring, minor bug fixes.
Updated renv.lock.

The single-sample pipeline now supports another two inputs:
- A delimited text file (table) representing a feature-barcode matrix.
- A SingleCellExperiment object saved as a Rds file.
Input type is controlled by a new INPUT_DATA parameter (replaces INPUT_10X_DIR) in 01_input_qc.yaml config. Details can be found in vignette("stage_input_qc").
Removal of empty droplets can be now disabled (EMPTY_DROPLETS_ENABLED parameter).
It is now possible to turn off normalization (NORMALIZATION_TYPE: "none") as the input can be an already normalized SingleCellExperiment object (meaning it went through the scdrake pipeline before).
Minor updates and bug fixes, updated documentation.

Fixed download_yq():
- download_yq() now supports the masOS (darwin).
- Fixed URL to Windows version (.exe extension).
- Better OS resolution implemented in .get_os().
Fixed issues with yq tool’s binary download on Windows.
Modified .confirm_menu() for non-interactive usage.
Activate an RStudio project on the end of init_project() function.
Fixed links to source code on GH.