Every parameter the pipeline accepts, grouped as in --help. They are checked against nextflow_schema.json before anything is downloaded or submitted. nextflow run main.nf --help <param> prints the same details for one parameter.
CSV file with columns id and path. path points to a CellRanger/STARsolo output directory (any CellRanger flavour), a .h5 file or a .mtx directory. With --on_irods it is an iRODS collection instead. See https://cellgeni.github.io/nf-cellbender/reference/samples/
--on_irods
boolean
false
Paths in --samples are iRODS collections; download them first.
--ignore_extensions
string
–
Comma-separated file extensions to skip when downloading from iRODS (with --on_irods only). Default: bam,cram,fastq,fq,fastq.gz,fq.gz,fastq.bz2,fq.bz2,fastq.xz,fq.xz,fastq.lz4,fq.lz4,mate1.bz2,mate2.bz2.
--outdir
string (directory)
results
Directory to save results.
CellBender options
Parameter
Type
Default
Description
--version
string or number
0.3
CellBender version. Allowed: 0.2, 0.3. Passing --version 0.2 on the command line gives a number, so both 0.2 and "0.2" are accepted.
--cells
integer
–
Expected number of cells (--expected-cells). Range: ≥ 1. Required together with --droplets for version 0.2 when the sample table points to .h5 files or .mtx directories. Otherwise version 0.2 estimates it with the mapper preset and version 0.3 with CellBender’s own estimation.
--droplets
integer
–
Total number of droplets to include (--total-droplets-included). Range: ≥ 1. Required together with --cells for version 0.2 when the sample table points to .h5 files or .mtx directories.
--mapper_preset
boolean
false
Estimate --cells, --droplets and --min_umi from the CellRanger/STARsolo output. Needs the whole mapper output directory as path. Always on for version 0.2 unless both --cells and --droplets are given. See https://cellgeni.github.io/nf-cellbender/guides/cell-calls/
--starsolo_mapper
string
GeneFull
STARsolo output type to use as CellBender input (e.g. Gene, GeneFull, GeneFull_Ex50pAS).
--exclude_features
string
–
Feature types to exclude from the analysis. Version 0.3: comma-separated list of Antibody Capture, CRISPR Guide Capture, Custom, Peaks, Multiplexing Capture, VDJ, VDJ-T, VDJ-T-GD, VDJ-B, Antigen Capture. Version 0.2: only All (excludes antibody capture). See https://cellgeni.github.io/nf-cellbender/guides/cell-calls/
--epochs
integer
–
Number of training epochs. Range: ≥ 1.
--fpr
number
–
Target false positive rate. Range: ≥ 0, ≤ 1.
--lr
number
–
Learning rate. Range: > 0.
--min_umi
integer
–
Lower bound for empty-droplet UMI count (--low-count-threshold). Range: ≥ 0.
--force_empty_umi_prior
integer
–
Force the empty-droplet UMI count prior (version 0.3 only). Range: ≥ 0.
QC detail level. Allowed: 1, 2, 3. 1: log-based summary and training curves only; 2: also cell calling, latent PCA and top ambient genes from the output .h5; 3: also reads the CellBender input for the UMI curve, soup fraction and per-gene removal.
Cluster options
Parameter
Type
Default
Description
--gpuqueue
string
gpu
LSF GPU queue for CellBender jobs. cub and tiger queues require --costcode. Allowed: gpu, cub22-inference, tiger22-inference. See https://cellgeni.github.io/nf-cellbender/guides/gpu-queues/
--costcode
string
–
Costcode to bill GPU jobs to (lowercased). Required for cub and tiger queues, not allowed with gpu.
--ignore_failed
boolean
false
Ignore CellBender jobs that still fail on the final retry instead of stopping the pipeline.
Rules across parameters
Combinations that fail validation, with the message you get:
--costcode is required for cub and tiger queues
--costcode is only used with cub and tiger queues
CellBender 0.2 needs --cells and --droplets together, or neither to use the mapper preset
CellBender 0.2 only supports --exclude_features All
--force_empty_umi_prior is only available for CellBender 0.3
--estimator is only available for CellBender 0.3
CellBender 0.2 uses the mapper preset automatically unless both --cells and --droplets are given