DO.CellBender

It is supposed to make something similar than the no longer working DietSeurat function, by removing no longer needed layers from th object. This function wraps a system call to a bash script for running CellBender on CellRanger outputs. It ensures required inputs are available and optionally installs CellBender in a conda env.

Usage

DO.CellBender(
  cellranger_path,
  output_path,
  samplenames = NULL,
  cuda = TRUE,
  cpu_threads = 15,
  epochs = 150,
  lr = 1e-05,
  estimator_multiple_cpu = FALSE,
  log = TRUE,
  conda_path = NULL,
  BarcodeRanking = TRUE,
  bash_script = system.file("bash", "_run_CellBender.sh", package = "DOtools")
)

Arguments

cellranger_path

Path to folder with CellRanger outputs.

output_path

Output directory for CellBender results.

samplenames

Optional vector of sample names. If NULL, will autodetect folders in cellranger_path.

cuda

Logical, whether to use GPU (CUDA).

cpu_threads

Number of CPU threads to use.

epochs

Number of training epochs.

lr

Learning rate.

estimator_multiple_cpu

Use estimator with multiple CPU threads (experimental).

log

Whether to enable logging.

conda_path

Optional path to the conda environment.

BarcodeRanking

Optional Calculation of estimated cells in samples through DropletUtils implementation

bash_script

Path to the bash script that runs CellBender.

Examples

if (FALSE) { # \dontrun{
# Define paths
cellranger_path <- "/mnt/data/cellranger_outputs"
output_path <- "/mnt/data/cellbender_outputs"

# Optional: specify sample names if automatic detection is not desired
samplenames <- c("Sample_1", "Sample_2")

# Run CellBender (uses GPU by default)
DO.CellBender(cellranger_path = cellranger_path,
              output_path = output_path,
              samplenames = samplenames,
              cuda = TRUE,
              cpu_threads = 8,
              epochs = 100,
              lr = 0.00001,
              estimator_multiple_cpu = FALSE,
              log = TRUE)
} # }