Easy installations and virtual environments with conda

Maxime Borry
24/01/2020

Follow the presentation

maxibor.github.io/conda-presentation

What does it mean to install a program ?

-> Copy/move the executable of the program in the `PATH`

What is the `PATH` ?

borry@mpi-sdag1:~$ echo $PATH
/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/
games:/usr/local/games:/opt/dell/srvadmin/bin

/usr/local/sbin
/usr/local/bin
/usr/sbin
/usr/bin
…

When can't “install” programms on a computer, it's because you don't have write access to these directories

What's in my path ?

For example, let's have a look at /usr/bin !

borry@mpi-sdag1:~$ ls -1 /usr/bin
...
head
headerdoc2html
heap
heap32
hexdump
hidutil
hiutil
host
...

Great, but what about Conda ?

I have a confession: this is not my real path

borry@mpi-sdag1:~$ echo $PATH
/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/
games:/usr/local/games:/opt/dell/srvadmin/bin

Here is my real path

borry@mpi-sdag1:~$ echo $PATH
/projects1/clusterhomes/borry/miniconda3/bin:/usr/local/
sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/
games:/usr/local/games:/opt/dell/srvadmin/bin

The location where executables will be installed by Conda

/projects1/clusterhomes/borry/miniconda3/bin

Conda

“Conda is an open source, cross-platform, language-agnostic package and environment management system distributed by Anaconda.”

Free
No admin rights required
Tons of bioinformatic packages available
Easy to install

Install conda

Two distributions:

Anaconda (fully featured, heavy)
Miniconda (bare minimum)

Installation for Mac

borry@maxime:~$ wget https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-x86_64.sh -O ~/miniconda.sh
borry@maxime:~$ bash ~/miniconda.sh

Installation for Linux

borry@mpi-sdag1:~$ wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda.sh
borry@maxime:~$ bash ~/miniconda.sh

Using conda

Install a package

When you install a package, conda automatically handles the installation of all its dependancies

Jupyter notebook

(base)borry@mpi-sdag1:~$ conda install jupyter

Install a package from a specific channel

BWA

(base)borry@mpi-sdag1:~$ conda install -c bioconda bwa

Bioconda: Channel for bioinformatics packages

Bioconda: sustainable and comprehensive software distribution for the life sciences

conda-forge: community-led (huge) collection of recipes

Everything with conda is an environment

An environement is an isolated sandbox that allows a fine control on program's versions and dependancies

By default, you're in the base environment

(base)borry@mpi-sdag1:~$ conda env list
# conda environments:
#
base          *  /projects1/clusterhomes/borry/miniconda3

But can create new environments…

(base)borry@mpi-sdag1:~$ conda create -n myEnvName

Everything with conda is an environment(2)

Change your current environemnt (activate)

(base)borry@mpi-sdag1:~$ conda activate myEnvName

Note that your prompt now includes the name of the active environment

Install a package in this environment, for example, RStudio

(myEnvName)borry@mpi-sdag1:~$ conda install -c r rstudio

And go back to the base environment

(myEnvName)borry@mpi-sdag1:~$ conda deactivate

The packages installed in one environement are not accessible from outside !

Let's create an environment for MetaPhlan2 and Krona

Krona plots

(base)borry@mpi-sdag1:~$ conda create -n metaphlan
(base)borry@mpi-sdag1:~$ conda activate metaphlan
(metaphlan)borry@mpi-sdag1:~$ conda install -c bioconda metaphlan2
# Now you can use metaphlan !
(metaphlan)conda install -c bioconda krona

Now let's share our environement !

(metaphlan)borry@mpi-sdag1:~$ conda env export > metaphlan_env.yml

And recreate it on another machine from the environment file

(base)borry@maxime:~$ conda env create -f metaphlan_env.yml

Other useful conda commands

List installed packages and versions in an environment

(metaphlan) borry@mpi-sdag1:~$ conda list

Uninstall a package (here Krona)

(metaphlan)borry@mpi-sdag1:~$ conda remove krona

Delete an environment

(base)borry@mpi-sdag1:~$ conda env remove -n metaphlan

Conda documentation: conda.io/docs

Faster conda with Mamba

Mamba is conda drop-in replacement, with a faster dependancy solver.

conda install -c conda-forge mamba