SalamiSnake_CLIP

Background

Snakemake pipeline for FLASH, eCLIP, ChIP-Seq and PARCLIP data. Changes needs to be done in order to adapt to your data. The file Snakefile includes evey step up to the point of the quality check of the mapping. The cluster.yml can be amended to give some tools more threads and memory. The config.yml need to be adapted for your data. The .qsub scripts can then be called with qsub *.qsub to queue it into SGE (Son of Grid Engine). Conda environments are placed into envs/. Other additional scirpts can be placed into scripts/. Rules that are imported into the Snakefile are located in rules/.

Input Data

Data needs to be provided in .fastqsanger and it must be de-multiplexed.

Config

To apply the pipleine to your data you need to specifiy the following things inseide the config.yml.

Define the names of your flash and control experiments:
- first define a key-name for each replicate (this is your choice)
- then insert the name of your fastqsanger files without the .fastqsanger extension
- files need to be correctly sorreted into r1 = forward and r2 = reverse order
- NOTE!: The key-name for the files needs to be consistent, i.e., the pipeline do not allow for something like: EBF1_rep1 and HR_EBF1_rep2. It need to be EBF1_rep1 and EBF1_rep2.
Define the path where the data is located in sample_data_dir:
Change output directory names if you want.

Qsub Files

You need to define the following things in the two qsub files:

Define the output and error log path (-o and -e)
Miniconda path MINICONDA=
Email to message when job fails -M and !! also in the snakemake command !!

Cluster.yml

You can specifiy the resources of various tools. You also need to define the output and error log path in out and err for each tool.

cores = number of threads
mem = memory

Conda environments

You can place conda environemnts inside envs. Take a look into the Snakemake guide for conda environments. To automatically install the tool itself in the conda environment, place the name of the tool inside the dependencies. The dependencies for the tools can be found in the conda github for recipies, or in the diretory of each tool in your miniconda.

Tools can be linked to a conda envs inside the Snakemake pipline with conda:.

Snakefile

The Snakefile is mainly responsible for the pre-processing of your data. It can be executed with the snakemake_qsub.sh script.

You need to define the following things:

Path to your reference genome REF_GENOME_DIR. Following data is needed:
- genome in fasta format
- transcriptome annotation in .gtf format
- chromosome sizes in .txt format
Path to your config.yml configfile
Adapter sequences inside the Cutadapt calls

Furthermore you need to keep in mind:

Activate got_umis and barcode_header_removal to extract UMIs and barcodes from your reads if not already done.
Activate estimate_insert_size to estimate the insert size if you are using PEAKachu for peak calling. It will give you information about this parameter.
Deactivate filter_out_unlocalized_regions_for_later_genome_versions if you are using versions of genomes which do not include these regions.

SnakefilePeakcalling.py

The SnakefilePeakcalling.py file is mainly responsible for the peak calling. It can be executed with the snakemake_peakcalling_qsub.sh script. The data needs to be pre-processed with the Snakefile beforehand.

You need to define the following things:

Path to your reference genome REF_GENOME_DIR. Following data is needed:
- genome in fasta format
- transcriptome annotation in .gtf format
- chromosome sizes in .txt format
Path to your config.yml configfile
When using PureCLIP only one side of the read pair is used. Based on the protocol change the samtools flag inside mate_reads_fitlering. Normally the crosslinking site is located on the forward (first) read in a standard FLASH protocol, thus the flag is 0x0040. Change it to 0x0080 for the reverse (second) read.

You can additionally change the following things:

Either use PureCLIP or PEAKachu for peak calling. When using PureCLIP activate the following things:
- mate_reads_fitlering
- pureclip
- intersect_binding_regions_with_peaks for PureCLIP

Further Remarks

Keep the following thins in mind:

Make yourself familiar with the tools.
Cutadapt is very important to change
FASTQC protocols are very important to look at.
PEAKachu Pairwise Replicates set it on true if all replicates from the signal and control come from the same sample. It models batch effects. true = (~ samples + conditions); false = (~ conditions)
PureCLIP works best with only one side of the paired reads.
PureCLIP works best if you only estimate the HMM parameters for three to four chromosomes. Else you get problewms with runtime and memory.
PureCLIP binding regions are often only one nucleotide long apart form the crosslinking site file. Further analysis require SlopBED to extend these regions.

heylf / SalamiSnake