Pipeline to create the multiscale methylation plot

Components of the workflow

1. Calculate x-values for plot
2. Add chromosome lengths to x-values file (needed for next step)
3. Create windows around x-values for calculating average methylation value
4. Calculate average methylation values for each window
5. Calculate cross-sample mean and standard deviation for each window size

Dependencies

The following dependencies are downloaded with --use-conda, otherwise you must have them in your PATH.

• snakemake (version 6.0+)
• bedtools
• python (version 3.7+)
  • pandas

Running the pipeline

• Clone or download the repo.
  • git clone [email protected]:huishenlab/multiscale_methylation_plot_pipeline.git, OR
  • Download from the releases page.
• Place gzipped BED files in the raw_data directory.
  • Alternatively, you can specify the path to your gzipped BED files in config/config.yaml.
• Replace the temporary sample names in config/samples.tsv with your sample names (everything before .bed.gz in your input files.
  • Alternatively, you can specify the path to your sample sheet in config/config.yaml. If you use your own file, be sure to include "sample" as the header line.
• Modify the config/config.yaml file with your chosen inputs.
  • At minimum, you will need to specify the FAI index file location for your genome. All inputs have defaults set.
• The pipeline can then be run on the command line (snakemake --cores 1 --use-conda) or submitted to a job scheduler on a cluster (a PBS script is provided: qsub bin/run_snakemake_workflow.sh).

After the pipeline finishes running

• Three directories will be created in your specified output directory.
  • analysis/: results from the pipeline
    • means/: mean methylation values in bins of varying size for each input sample
    • stats/: average and standard deviation for values in each bin across all samples
    • x_values/: bins used for calculating mean values
  • benchmarks/: benchmarking data compiled when each rule runs
  • logs/: log files generated as rules finish

Creating the multiscale plot

The multiscale plot can be created using bisplotti, specifically the multiscaleMethylationPlot() function. Input is a specific sample in analysis/means/ or the average for all samples from analysis/stats/. If you have two (or more) sample groups that you processed together, you can also find per-bin average values for a specified set of samples using multiscaleGroupAverage() in bisplotti.

Example dataset

An example dataset has been provided as a .zip file on the releases page. Instructions for using can be found in a README once the file has been unzipped.