Installation

Using bioconda (recommended)

The recommended way to install IM-Fusion is using bioconda, as this installs IM-Fusion together with its external dependencies into an isolated environment using a single command:

conda create -n imfusion imfusion

This assumes that condas channels have been configured as recommended by bioconda.

Alternatively, IM-Fusion can be installed in an existing environment using:

conda install imfusion

Note that the bioconda package does not install rpy2, which is only required for the single-sample differential expression (DE) test and can be difficult to install. If you wish to use the single-sample DE test, you can install R and rpy2 can be installed using as follows:

# Using conda
conda install rpy2

# Using pip
pip install rpy2

Using pip

IM-Fusion can also be installed using pip:

pip install imfusion

Note that this does not install any of the required external dependencies, which must therefore be installed manually (see below for more details). This does also not install rpy2, which must also be installed using pip if you plan to use single-sample differential expression test.

If you don’t already have numpy installed, you may run into the following error during installation:

Downloading HTSeq-0.7.2.tar.gz (548kB)
    Complete output from command python setup.py egg_info:
    Setup script for HTSeq: Failed to import 'numpy'.
    Please install numpy and then try again to install HTSeq.

This can be fixed by first installing numpy before trying to install IM-Fusion, as numpy is an install-time dependency of one of IM-Fusions dependencies (HTSeq). Additionally, HTSeq requires liblzma to be installed to build properly, otherwise you will encounter the following error:

/usr/bin/ld: cannot find -llzma
collect2: error: ld returned 1 exit status

This can be fixed by installing the liblzma-dev package. On linux (assuming an apt-get based distribution) this can be done as follows:

sudo apt-get update
sudo apt-get -y install liblzma-dev

Hardware dependencies

To build references and align reads using the STAR aligner, you will need to have ~30GB RAM. For analyses with Tophat2, you will typically need ~5-15GB of RAM, depending on the number of reads in your sample.

Software dependencies

IM-Fusion has been tested on macOS and linux, and is compatible with both Python 2.7 and Python 3.4+. However, identifying insertions using Tophat-Fusion only works in Python 2.7, as Tophat2 is unfortunately not compatible with Python 3.

IM-Fusion has the following external dependencies that must be installed for full functionality:

• STAR - for insertion detection using STAR.
• Tophat2/Bowtie (Python 2.7 only) – for insertion detection using Tophat-Fusion.
• featureCounts – for generating expression counts.
• StringTie (optional) - for performing reference-guided transcript assembly.
• STAR-Fusion (optional) – for identifying endogenous gene fusions using STAR.
• Sambamba (optional) – for faster sorting of BAM files.

Note that either STAR or Tophat2/Bowtie need to be installed (not both necessarily), depending on your preference for the RNA-seq aligner. If you are unsure, we recommend using STAR. StringTie is only required if you plan to perform reference-guided assembly (used to detect novel transcripts resulting from insertions). Similarly, STAR-Fusion is only needed if you plan to do endogenous fusion detection. Sambamba is used for faster sorting of BAM files in the STAR pipeline, but is also not a strict requirement.

Each of these packages can be installed manually by downloading binaries from the respective websites and making sure the unpacked binaries are available in the current $PATH. Alternatively, the dependencies can easily be installed using bioconda:

conda install -c bioconda star bowtie tophat stringtie \
    subread star-fusion sambamba

Note that Tophat2 will not function properly in Python 3. Bioconda also does not provide macOS builds of STAR-Fusion, meaning that STAR-Fusion will have to be installed manually on macOS.

If you don’t use conda, a number of these tools can also be installed using homebrew or linuxbrew:

brew tap homebrew/science
brew update
brew install bowtie star stringtie tophat sambamba

Unfortunately, STAR-Fusion and subread (featureCounts) are currently not available in homebrew and will therefore need to be installed manually.