Transcription Start Site HMM for GROcap data

doi:10.1038/ng.3142

Dependencies

This code requires R and is only know to build on Linux and OS X.

The R scripts in the 'pipeline' folder require the following R packages from CRAN:

• parallel
• grid
• Roxygen or Roxygen2

Additionally, the following R packages, which can be found in the "rpkg" folder, must be built from source

• bigWig
• QHMM
• grocaptss

Setup

Start by clonning this repository with it's submodules:

git clone --recursive https://github.com/andrelmartins/grocap.tsshmm

After cloning this repository, and assuming you have installed the pre-requisite CRAN packages, go into the "rpkg" folder and run make. This will build the required source R packages and install them in your system.

Datasets

This pipeline requires the following datasets (in bigWig format, one bigWig file per strand):

• GRO-cap TAP+
• GRO-cap TAP-
• Nuclear CAGE PolyA Plus (single base pair resolution pileups, not the smoothed versions found in the UCSC Genome Browser)
• ChromHMM Predictions (for post-processing statistics)

Running

• define your dataset (see common.R and rpkg/grocaptss/pkg/R/dataset.R)
• compute normalization totals for both the main dataset (TAP+) (example: data/gm12878.totals.norm.Rdata) and background (TAP-) (example: data/gm12878.totals.norm.back.Rdata) (see pipeline/hmm.parse.gm12878.R)
• create your copy of the pipeline scripts pointing to your dataset
• execute the pipeline steps in turn:
  • hmm.parse (example: Rscript hmm.parser.gm12878.R)
  • postproc
  • postproc.stats [optional]
  • classify