Ambient RNA Correction with CellBender

Author

Ahmed M. Elhossiny

The first step in the workflow, prior to analysis in R, is to correct the raw count matrices for ambient RNA contamination. This is performed using CellBender [@cellbender]

Environment Setup

The exact conda environment used in this run can be installed using this yml file using conda env create -f cellbender.yml

Alternatively you can create the environement as follows

## Installing cellbender
conda create -n cellbender python=3.7
conda activate cellbender
pip install cellbender==0.3.0

Running Cellbender

The following shell script is submitted to a high-performance computing (HPC) cluster as a SLURM job array to process each of the 48 samples in parallel.

The script pulls a cellbender_manifest.txt file from data directory with at two columns: sample_id and cellranger_output path for each sample as follow

sample_1	path/to/sample_1/cellranger_output
sample_2	path/to/sample_2/cellranger_output
sample_3	path/to/sample_3/cellranger_output

#!/bin/bash

#SBATCH --account=
#SBATCH --job-name='Cellbender_scRNAseq_%a'
#SBATCH --output=logs/Cellbender_scRNAseq_%a.log
#SBATCH --partition=gpu
#SBATCH --mem=128G
#SBATCH --cpus-per-task=16
#SBATCH --gres=gpu:1
#SBATCH --time=04:00:00
#SBATCH --array=1-48

## Setting up environment
echo -e ">>> Start time $(date) <<<"
start_time=$(date +%s)

source activate cellbender
outputDir=../outputs/scRNASeq_Analysis/cellbender/
mkdir -p $outputDir

## Importing samples info
samples_info=../data/cellbender_manifest.txt
sample=$(cat $samples_info | cut -f1 | sed -n $[SLURM_ARRAY_TASK_ID]p)
path=$(cat $samples_info | cut -f2 | sed -n $[SLURM_ARRAY_TASK_ID]p)
outputDir=${outputDir}/${sample}
mkdir -p $outputDir
echo -e ">>> Processing $sample <<<"
echo -e ">>> Sample path $path <<<"
echo -e ">>> Output in $outputDir <<<"

## Running CellBender 
cd $outputDir
cellbender remove-background \
--cuda \
--input ${path}/raw_feature_bc_matrix.h5 \
--output ${outputDir}/${sample}.h5

## Reporting time
echo -e ">>> End time $(date) <<<"
end_time=$(date +%s)
runtime_seconds=$((end_time - start_time))
runtime_minutes=$((runtime_seconds / 60))
echo "Total runtime: $runtime_minutes minutes"

Next Steps

With Cell Ranger outputs, proceed to scRNAseq Processing & Integration