/bedshift_analysis

Primary LanguagePython

bedshift analysis

Contents of the repository

This repository contains results from Bedshift: perturbation of genomic interval sets, a paper describing application of the bedshift tool to explore the behavior of region set similarity metrics on simulated data.

There are 4 experiment folders. Each of these is organized as a PEP that includes a CSV file containing bedshift parameter sets:

  • pep_demo provides a quick demo that runs quickly for testing purposes.
  • pep_basic provides a set of 36 parameter sets to test combinatorial perturbations on a single input file
  • pep_main runs the same parameter sets, but across 3 different input files
  • pep_universe runs the same parameter sets with the original file from pep_basic, but using 3 different universes

For each experiment, you will find a table of perturbations, with one row per perturbation, and a configuration file. For example, in the pep_main project, look at sample_table.csv. The config file points to this.

This repository also contains 2 pipeline interface files the piface_bedshift.yaml, describes how to run bedshift on the specified bed file with the listed perturbation parameters; and the piface_similarity_scores.yaml file runs our included similarity score calculation script (in src).

Setup and configuration

Refgenie setup

To use refgenie to grab the chrom sizes file, just do:

pip install refgenie
export REFGENIE="refgenie_config.yaml"
refgenie init -c $REFGENIE
refgenie pull hg38/fasta

Environment setup

export CODE=/path/to/directory

Where directory is the directory containing this repo, bedshift_analysis.

Demo

Run bedshift

Start with the PEP in the pep_demo folder. Everything needed to run this is stored in this repository, and it's a short example to show that you have everything set up correctly. Run it like this.

looper run pep_demo/project_config.yaml

Or, to use bulker, just do looper run pep_demo/project_config.yaml -p bulker_local. This will produce an output folder in results/pep_demo/bedshifted_regions.

Aggregate bedshift scores

After this is complete, you can aggregate the results and generate the scores with runp like this:

looper runp project_config.yaml

Or looper runp project_config.yaml -p bulker_local. This will create score files in pep_demo/results/scores.

Make plots

Once the scores are generated, you can reproduce the plots we made in the paper by following the R script in src/plot_summary_results.R.

Real project

There are 3 real projects in here:

  • pep_basic is a basic run, which tests a single BED file against a single universe
  • pep_main is the main project, which tests 3 BED files, using just 1 universe.
  • pep_universe tests how different universes behave, using 1 BED file and 3 different universes.

All projects can be run the same way.

Download data

Run ./src/download_data.sh to download all the data required to run all projects.

The bed files will be downloaded from bedbase:

The universes are downloaded from SCREEN:

You can also see more information about the universes at bedbase.org:

Run bedshift

Here are some example looper run commands:

looper run pep_main/project_config.yaml -p local
looper run pep_universe/project_config.yaml -p local
looper run pep_universe/project_config.yaml --lumpn 10
looper run pep_universe/project_config.yaml --lumpn 10 --sel-attr trial --sel-incl 0 1 2

Aggregate results

looper runp pep_universe/project_config.yaml -l 1 -p local
looper runp pep_main/project_config.yaml