epi-refs-smk_v8

Rational

• Epi-refs-smk_v8 is a basic and for sure to simplified snakemake pipeline to quickly generate epigenomic references for downstream projects.
• Installation of the base snakemake v8 conda env: conda env create -f base_smk_8v.yml
• Add the epigenomic references here: config/sample_manifest.tsv
• This pipeline is very conservative in processing and instead of adding a pseudocount it rather drops sites with insufficient coverage.

To dos

• Implement peak calling
  • Peak calling uses macs3 and currently only works when config['AGGREGATE_REPLICATES] == 'sum'
  • Also currently only works for "single end" reads, independend whether SE sequenced or PE sequenced and only one read is retained for analysis
• Implement a window based analysis
• Call chromatin states using ChromHMM or another model

Set up

1. Install the base snakemake env
2. Fill out the config/sample_manifest.tsv

• Read the config/sample_manifest_notes.md

3. Set the corresponding reference genome in config['REFERENCE_GENOME]

• Check whether it is available at UCSC if not:
• Set the config['USE_CUSTOM_GENOMES'] to True and specify the config['CUSTOM_GENOME_PATH']
• CUSTOM_GENOME_PATH can be a local or remote file, the pipeline also handles compressed or uncompressed reference genomes.

3. Enable / disable whether both reads of a paired read sequencing experiment should be considered

• config['PE_SEQ_USE_SINGLE_READ']
• specify: 1 for read 1, or 2 for read 2, or leave empty if not entended
• The pipeline will align both reads but then only considers one read in the downstream processing after the PCR deduplication

4. Enable / disable the usage of the blacklists with config['REMOVE_BLACKLISTED_REGIONS']

• To be as flexible as possible one can switch on / off the usage of precomputed blacklists.
• Keep in mind it is likely when using custom genomes (config['USE_CUSTOM_GENOMES']) no blacklist exist, therefore it should be disabled.

5. Enable / disable the autosomal chromosome filtering config['FILTER_AUTOSOMAL_CHRS']

• Same reasoning as 4. custom genomes might have different chromosome names and therefore autosomal filtering will lead to wrong exclusion of chrs. Currently filtering assumes chromosome names are: chr1... chr20

Workflow overview