ATAC-seq snakemake pipeline

Snakemake pipeline for processing ATACseq PE data.

How to use

1. Prepare work environment

• Clone this repository in the working directory

git clone https://github.com/luostrowski/atacseq_snakemake.git

• Navigate into the data/fastq directory

cd data/fastq

• Symlink files to the directory

ln -s /absolute/path/to/file_R1.fq.gz .
ln -s /absolute/path/to/file_R2.fq.gz .

• Rename symlink to proper format

mv file_R1.fq.gz sample_R1.fastq.gz
mv file_R2.fq.gz sample_R2.fastq.gz

2. Adjust configuration file

The proj_config.yaml file contains the wildcards that the script will use.

• Change the path to the reference genome (it must be indexed).

• Change the path to the annotation gtf file.

• Adjust treat samples if necessary.

• Change trimmomatic adapters if necessary. Default is Nextera PE, usually works with ATAC-seq.

• Change the mitochondrial genome name (mt_chr) if necessary. Options are: "chrM" or "MT".

• Adjust effective_genome_size for deeptools bamcoverage:
  - GRCh37: 2864785220
  - GRCh38: 2913022398
  - GRCm37: 2620345972
  - GRCm38: 2652783500

• Adjust other parameters if necessary:

  • macs2_genome: two letter code for MACS2 to recognize genome
  • contrasts: desired contrasts for differential analysis
  • analysis_method: for diffbind, either DBA_DESEQ2 or DBA_EDGER
  • min_overlap: for diffbind merging, min number of samples peak found to retain it
  • fdr_cutoff: padj cutoff for defining significant diff peaks
  • data_source: for chipseeker, genome used
  • organism: for chipseeker, organism latin name with underscore (no spaces)

4. Execute the snakefile

• First try a dry run to make sure the structure works.

snakemake --dry-run

• Adjust the cluster.json file to run on a cluster submitting slurm jobs.

  • The --latency-wait 60 parameter is required when using slurm to make sure outfiles are complete.

snakemake --use-conda --jobs 100 --latency-wait 60 --cluster-config cluster.json --cluster "sbatch --qos {cluster.qos} -p {cluster.partition} -N {cluster.nodes} -n {cluster.cores} --mem {cluster.mem} -t {cluster.time} -o {cluster.stdout} -e {cluster.stderr}"

• If it is necessary to re-run the script, the incomplete files will need to be re-generated.

snakemake --rerun-incomplete --use-conda --jobs 100 --latency-wait 60 --cluster-config cluster.json --cluster "sbatch --qos {cluster.qos} -p {cluster.partition} -N {cluster.nodes} -n {cluster.cores} --mem {cluster.mem} -t {cluster.time} -o {cluster.stdout} -e {cluster.stderr}"

Directed acyclic graph of the steps