#RColumbo

Installing RColumbo (without root privelages) (on RCC)

0. Load required modules and edit .bash_profile

module load cmake

module load R

Add this line to the end of ~/.bash_profile:

export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH

then run source ~/.bash_profile

1. Download and install HDF5

The HDF5 module on RCC does not support C++

wget https://www.hdfgroup.org/ftp/HDF5/releases/hdf5-1.10/hdf5-1.10.0-patch1/src/hdf5-1.10.0-patch1.tar.gz

tar -xzf hdf5-1.10.0-patch1.tar.gz

cd hdf5-1.10.0-patch1

./configure --prefix=$HOME --enable-cxx --enable-build-mode=production

make install

2. Download blosc

git clone https://github.com/Blosc/c-blosc

cd c-blosc

mkdir build

cd build

cmake -DCMAKE_INSTALL_PREFIX=$HOME ..

make

make install

3. Install RColumbo

from within R #Install devools if necessary

install.packages(devtools)

library(devtools)

#Install rhdf5 if necessary

source("https://bioconductor.org/biocLite.R")

biocLite("rhdf5")

devtools::install_github("CreRecombinase/h5")

devtools::install_github("CreRecombinase/RColumbo",ref="release")

library(RColumbo)

4. Read compressed HDF5 files really fast

The eQTL results are stored as dataframes in HDF5 (.h5) files. There is one file for each chromosome. Each h5 file has a cis_eQTL dataframe and a trans_eQTL dataframe. The dataframes can be read using RColumbo's read_h5_df function:

library(RColumbo)

eqtl_file <- "/project/xinhe/eQTL/WB_Chr2_v6p_ortho_flip.h5"

cis_eqtl_df <- read_h5_df(eqtl_file,groupname="cis_eQTL")

trans_eqtl_df <- read_h5_df(eqtl_file,groupname="trans_eQTL")

The columns of these dataframes are as follows

• chrom: chromosome
• fgeneid: a gene identifier (it's a basically the ensembl gene ID)
• pos: chromosome coordinate as given by GTEx
• theta:estimated eQTL regression coefficient (no intercept)
• serr: standard error of eQTL regression coefficient