MUSiCC is a marker genes based framework for metagenomic normalization and accurate profiling of gene abundances in the microbiome, developed and maintained by the Borenstein group at the University of Washington.

MUSiCC is available through the following sources:

• As a Python module from GitHub or PyPI (see installation instructions below)
• As an online tool at: http://elbo.gs.washington.edu/software_musicc.html.

MUSiCC is distributed under a BSD license and can be readily incorporated into custom analysis tools.

Prerequisites for installing:

In order for MUSiCC to run successfully, the following Python modules should be pre-installed on your system:

• Numpy >= 1.6.1 (http://www.numpy.org/)
• Scipy >= 0.9 (http://www.scipy.org/)
• Scikit-learn >= 0.15.2 (http://scikit-learn.org/stable/)
• Pandas >= 0.14 (http://pandas.pydata.org/)

If you have pip installed, you can install these packages by running the following command:

pip install -U numpy scipy scikit-learn pandas

Installing MUSiCC:

To install MUSiCC, download the package from https://github.com/omanor/MUSiCC/archive/1.0.tar.gz

After downloading MUSiCC, you’ll need to unzip the file. If you’ve downloaded the release version, do this with the following command:

tar -xzf MUSiCC-1.0.tar.gz

You’ll then change into the new MUSiCC directory as follows:

cd MUSiCC-1.0

and install using the following command:

python setup.py install

ALTERNATIVELY, you can install MUSiCC directly from PyPI by running:

pip install -U MUSiCC

Note for windows users: Under some windows installations, Scipy may fail when importing the Stats module. Workarounds may be found online, such as here.

After downloading and installing the software, we recommend testing it by running the following command:

test_musicc.py

This will invoke a series of tests. A correct output should end with:

Ran 3 tests in X.XXXXs

OK

The MUSiCC module handles all calculations internally. MUSiCC offers an interface to the MUSiCC functionality via the command line and the run_musicc script.

run_musicc.py input_file [options]

input_file

Input abundance file to correct

-h, --help

show help message and exit

-o OUTPUT_FILE, --out OUTPUT_FILE

Output destination for corrected abundance (default: MUSiCC.tab)

-if {tab,csv}, --input_format {tab,csv}

Option indicating the format of the input file (default: tab)

-of {tab,csv}, --output_format {tab,csv}

Option indicating the format of the output file (default: tab)

-n, --normalize

Apply MUSiCC normalization (default: false)

-c {use_generic, learn_model}, --correct {use_generic,learn_model}

Correct abundance per-sample using MUSiCC (default: false)

-perf, --performance

Calculate model performance on various gene sets (may add to running time) (default: false)

-v, --verbose

Increase verbosity of module (default: false)

MUSiCC can also be used directly inside a python script. Passing variables and flags to the MUSiCC script is done by creating a dictionary and passing it to the function correct_and_normalize, as shown below.

>>> from musicc.core import correct_and_normalize >>> musicc_args = {'input_file': 'test_musicc/lib/python3.3/site-packages/musicc/examples/simulated_ko_relative_abundance.tab', 'output_file': 'MUSiCC.tab','input_format': 'tab', 'output_format': 'tab', 'musicc_inter': True, 'musicc_intra': 'learn_model','compute_scores': True, 'verbose': True} >>> correct_and_normalize(musicc_args)

input_file

Input abundance file to correct

output_file

Output destination for corrected abundance (default: MUSiCC.tab)

input_format {'tab','csv'}

Option indicating the format of the input file (default: 'tab')

output_format {'tab','csv'}

Option indicating the format of the output file (default: 'tab')

musicc_inter {True, False}

Apply MUSiCC normalization (default: False)

musicc_intra {'use_generic', 'learn_model', 'None'}

Correct abundance per-sample using MUSiCC (default: 'None')

compute_scores {True, False}

Calculate model performance on various gene sets (may add to running time) (default: False)

verbose {True, False}

Increase verbosity of module (default: False)

In the musicc/examples directory, the file simulated_ko_relative_abundance.tab contains simulated KO abundance measurements of 20 samples described in the MUSiCC manuscript. Using this file as input for MUSiCC results in the following files:

• simulated_ko_MUSiCC_Normalized.tab (only normalization)
• simulated_ko_MUSiCC_Normalized_Corrected_use_generic.tab (normalize and correct using the generic model learned from HMP)
• simulated_ko_MUSiCC_Normalized_Corrected_learn_model.tab (normalize and correct learning a new model for each sample)

The commands used were the following (via command line):

run_musicc.py musicc/examples/simulated_ko_relative_abundance.tab -n -perf -v -o musicc/examples/simulated_ko_MUSiCC_Normalized.tab

run_musicc.py musicc/examples/simulated_ko_relative_abundance.tab -n -c use_generic -perf -v -o musicc/examples/simulated_ko_MUSiCC_Normalized_Corrected_use_generic.tab

run_musicc.py musicc/examples/simulated_ko_relative_abundance.tab -n -c learn_model -perf -v -o musicc/examples/simulated_ko_MUSiCC_Normalized_Corrected_learn_model.tab

If you use the MUSiCC software, please cite the following paper:

MUSiCC: A marker genes based framework for metagenomic normalization and accurate profiling of gene abundances in the microbiome. Ohad Manor and Elhanan Borenstein. Genome Biology

For MUSiCC announcements and questions, including notification of new releases, you can visit the MUSiCC users forum.