Before starting this section, please make sure that your data is preprocessed as described in the Dataset preparation section of this wiki.

Pipeline execution

The pipeline can be used with the following command:

nextflow run Mye-InfoBank/scRAFIKI -resume -profile <YOUR_PROFILE> --samplesheet "samplesheet.csv"

Samplesheet

The samplesheet is a csv file with the following column names as header:

Column	Description	Default
id	The dataset identifier, needs to be unique across datasets	required
input_adata	Path to the input AnnData (.h5ad) file	required
min_counts	Minimum number of counts per cell	0
max_counts	Maximum number of counts per cell	Infinity
min_genes	Minimum number of genes per cell	0
max_genes	Maximum number of genes per cell	Infinity
max_pct_mito	Maximum percentage of mitochondrial genes per cell	100
no_symbols	Set to true to enable conversion of var_names to gene symbols. If enabled, ensemble and entrez IDs can be processed.	false
transfer	Perform integration using the transfer learning method scArches	false

Cell filtering will be handled by the pipeline if the respective columns are specified. If filtering has been done before, the columns can be omitted.

Parameters

Parameter	Description	Default
samplesheet	Path to the samplesheet	required
outdir	Path to the output directory	./output
celltypist_model	Celltypist model to use for annotation, a list of possible values can be found here. Make sure to add the .pkl extension, e.g. Cells_Intestinal_Tract.pkl. If the parameter is null, celltypist will not be executed.	null
leiden_resolutions	List of resolutions for clustering	[0.25, 0.5, 0.75, 1, 1.5, 2]
integration_methods	List of integration methods to use	["scvi", "scanvi", "harmony", "scgen", "scanorama", "bbknn", "desc", "combat", "trvaep"]
scshc	Run sc-SHC clustering and QC	false
entropy	Calculate entropy of clusterings compared to batches	false
cell_cycle	Run cell cycle analysis	true
benchmark_hvgs	Number of highly variable genes to use for scIB benchmarking. Benchmarking will only be executed, if a positive value is set.	0
integration_hvgs	Number of highly variable genes to use for integrations.	6000
normalization_method	Method to use for normalizing count values. Can be one out of: raw (no normalization), scTransform, log_total, pearson_residuals	log_total
min_cells	Minimum number of cells that a gene needs to be expressed in so that it will be included in the pipeline output	10
decontX	Perform ambient RNA removal using decontX	true
has_extended	Set to true if any of the datasets has the transfer option activated in the samplesheet	false
has_celltypes	Set to false if none of the input datasets has annotations in the cell_type metadata field	true
custom_metadata	Custom metadata fields that should be passed through the pipeline - if not present in all datasets, Unknown will be used as a filler	[]
custom_hvgs	Can be used to manually add genes to the automated selection of highly variable genes	[]
publish_mode	Defines if the output files are copied (copy) or symlinked (symlink)	symlink
max_cpus	Maximum number of CPUs to use	60
max_memory	Maximum amount of memory to use	300.GB
max_time	Maximum runtime of a job	48.h

Parameters can be applied as documented in the Nextflow documentation.

Note that the integration methods scvi and scanvi will be executed regardless of the specified integration methods, as they are required for some downstream analysis steps.

Profiles

The following profiles are available:

• standard: Run the pipeline on your local machine without containerization, requires all dependencies to be installed (not recommended)
• docker (if running on an ARM-based machine, use docker,arm)
• singularity
• podman
• shifter
• charliecloud
• apptainer
• gitpod