Software setup for Practical GWAS using Linux and R

We use virtual machine (VM), e.g.,, in a distribution called GWAS-2017.dvi, which includes many pre-installed Linux software and R packages. It is available from

1. Oracle VirtualBox attachment

The Oracle VirtualBox software can be downloaded from

The latest stable version is 5.1.28. To install, under VirtualBox binaries select the appropriate host inside VirtualBox 5.1.28 platform packages.

Once installed and at the very first run, it will prompt for installation of VM VirtualBox Extension Pack. Follow instructions for a successful installation. From VirtualBox, select New, for a 64-bit Linux, Ubuntu, ~3GB RAM with an existing disk which points to GWAS-2017.vdi.

Once the disk is attached, you can start the VM to login with the following details,

username: physalia

password: course

If you run into difficulty, the following links are helpful,

In general with VirtualBox, if you don’t see full-screen you can follow these steps,

1). Make sure from VirtualBox menu, click Devices pull-down menu and select Insert Guest Addition CD Image and also Shared Clipboard, Bidirectional.

2). Once the desktop is up, right click the background to create a terminal, and perform the following,

cd /media/physalia/VBOXADDITIONS_5.1.28_117968/
sudo ./

3). If successful, type


When login again, you would be guaranted to have a full screen so you can start another terminal.

2. Notes for existing Linux users

While it is relatively straightforward to download the relevant packages as listed in /home/physalia/Downloads/www, it would still take some effort to make the relevant R packages, namely, ~200 packages at /usr/local/lib/R/site-library, in place.

To save time, you can also move the downloads over to your own.

3. Notes for Windows users

It is possible to set up a native Ubuntu system inside Windows 10 but in this case a major drawback is the difficulty for file exchange and use of all available resources.

From a terminal, issue the following command,


then you can see files on your C:\ from Ubuntu.

4. Notes for users of other operating systems

Similar procedures described above would work, but you would need create an appropriate sharefolder from the VM settings to enable file exchange.

5. Additional notes for installation

When the executable files are directly available, for instance QUICKTEST we can simply do

 sudo cp quicktest /usr/local/bin

This also applies to BOLT-LMM, PLINK and many others.

The following gives more details regarding other cases.

bcftools, samtools and vcftools

All can be installed into /usr/local/bin via sudo make install from their directories.


It requires java to be installed, e.g.,

sudo apt install default-jre

From beagle_4.1/, issue

chmod +x run.beagle.08Jun17.d8b.example

will also obtain additional files to run through the examples.


While make will generate the executable, the perl/ and python/ extensions are made available as follows,

perl Makefile.PL
sudo make install


sudo apt install python-dev
python build
sudo python install



As BLAS and LAPACK have been installed, we could make a copy of ` into followed by the usual make`.


Our copy was obtained from GitHub, The installation involves the following.

sudo apt install aclocal doxygen texinfo bison help2man
sudo apt install autoconf
sudo apt install automake
autoreconf --install
./configure --prefix=/usr/local
make install

The first line was suggested by ./missing --help and at least doxygen and texinfo are helpful.


The download is a .zip file, but the OpenBLAS provided does not work so we do

unzip PASCAL
sed 's/xianyi-OpenBLAS-48f06dd/xianyi-OpenBLAS-6d2da63/g' >

where we download the latest version 0.2.20 from SourceForge and use a revised installation script taking this into account. We still have to do one more thing, the issues are to do with

ln -s
ln -s

and we simply replace -s with -sf and bash again, saying (N)one when prompting for unzip OpenBLAS again.


This is furnished with the following steps,

tar jxf depict_140721.tar.bz2

pip2.7 install intervaltree
pip2.7 install pandas

This compressed tar ball also has the associate databases nevertheless its formidable 3GB size prevents us to pre-install them. Latest update is available from GitHub which would need invoke Java IDE such as NetBeans to compile.

Oddly, the example provided by the software only works upto four tissues it would need some guesswork to use the whole reconstitute genesets and our configuration file has already done so. We also used the database reported in the DEPICT paper.


The last section is heavy-going nevertheless critical. It avoids packaging large software such as DEPICT and then installing into our .vdi, especially volume of other software (not necessarily with their associate databases) reaches 20GB.