Cancer Systems Biology, Technical University of Denmark, 2800, Lyngby, Denmark
Cancer Structural Biology, Danish Cancer Institute, 2100, Copenhagen, Denmark
Repository associated to the publication:

MAVISp: A Modular Structure-Based Framework for Genomic Variant Interpretation Matteo Arnaudi, Ludovica Beltrame, Kristine Degn, Mattia Utichi, Simone Scrima, Pablo Sanchez Izquierdo, Karolina Krzesinska, Francesca Maselli, Terezia Dorcakova, Jordan Safer, Alberte Heering Estad, Katrine Meldgard, Philipp Becker, Julie Bruun Brockhoff, Amalie Drud Nielsen, Valentina Sora, Alberto Pettenella, Jeremy Vinhas, Peter Wad Sackett, Claudia Cava, Anna Rohlin, Mef Nilbert, Sumaiya Iqbal, Matteo Lambrughi, Matteo Tiberti, Elena Papaleo. bioRxiv https://doi.org/10.1101/2022.10.22.513328

This is the web app of the MAVISp database for structural variants annotation.

If you use MAVISp, please cite our preprint:

MAVISp: Multi-layered Assessment of VarIants by Structure for proteins
Matteo Arnaudi, Ludovica Beltrame, Kristine Degn, Mattia Utichi, Pablo Sánchez-Izquierdo, Simone Scrima, Francesca Maselli, Karolina Krzesińska, Terézia Dorčaková, Jordan Safer, Katrine Meldgård, Julie Bruun Brockhoff, Amalie Drud Nielsen, Alberto Pettenella, Jérémy Vinhas, Peter Wad Sackett, Claudia Cava, Sumaiya Iqbal, View ORCID ProfileMatteo Lambrughi, Matteo Tiberti, Elena Papaleo biorxiv, doi: https://doi.org/10.1101/2022.10.22.513328

Please see the CHANGELOG.md file on this repository for current and expected data releases as well as updates on the software.

Please see the CURATORS.md file in this repository for an up to date list of our curators.

This web app uses by default a database based on a set of CSV files, available on our OSF repository. These were generated by using the MAVISp Python package from raw data files. See below for details about the Python package.

Running the MAVISp web app requires a working Python 3.9+ installation with the following Python packages:

• streamlit 1.28.2
• streamlit-aggrid 0.3.4.post3
• pandas 2.1.3
• matplotlib 3.7.4

In principle, it is compatible with all operating systems that support Python.

It has been last test on Linux (Ubuntu 18.04), and on macOS (13.5.2), with Python 3.9.6 and the following package versions:

• streamlit 1.28.2
• streamlit-aggrid 0.3.4.post3
• pandas 2.1.3
• matplotlib 3.7.4

In order to download the full MAVISp dataset from OSF, you will also need the wget program (see below). We last tested the download with wget 1.21.3.

These instructions apply to both Linux and macOS, using the terminal. You will need to have a recent (>=3.9) Python distribution installed on your system or Anaconda.

1. if you have access to the virtualenv Python environment module, you can create a virtual environment:

virtualenv -p python3.9 MAVISp_env

2. then, activate it:

source MAVISp_env/bin/activate

3. you can install the requirements in the environment using pip:

pip install pandas==2.1.3 matplotlib==3.7.4 streamlit==1.28.2 streamlit-aggrid==0.3.4.post3

1. if you have access to Anaconda or Miniconda (executable conda), you can use it to create a virtual environment:

conda create -n MAVISp_env python

2. then you can activate it:

conda activate MAVISp_env

3. you need to install the remaining requirements, using pip:

pip install pandas==2.1.3 matplotlib==3.7.4 streamlit==1.28.2 streamlit-aggrid==0.3.4.post3

Installation time is typically up to a few minutes.

In the following instructions

• hostname denotes the hostname of the server (i.e. the one you usually ssh to)
• user denotes your username on the server

This requires downloading the full MAVISp dataset.

In order to run our web server locally with its full content, you will need to download the full MAVISp dataset from OSF, as follow, as well as download our web app from GitHub. If you'd rather test the web app on a small subset, please follow the instructions in the "Running the app locally - test dataset" instead.

1. If you haven't already, activate your Python environment (see previous steps)

2. create a local copy of the MAVISp repository in your system:

git clone https://github.com/ELELAB/MAVISp

3. Download the database files from our OSF repository. This requires the wget program or similar. If it's not available, you can manually download a zip file containing all the database files from this link.

cd MAVISp
rm -rf ./database
mkdir database 
cd database
wget -O database.zip 'https://files.de-1.osf.io/v1/resources/ufpzm/providers/osfstorage/65579865874c2e15e54e7d34/?zip=' && unzip database.zip 
rm database.zip
cd ..

At the end of the process, you should have a database folder inside the MAVISp folder including all the contents of the database folder on OSF.

4. With your Python environment still active and from inside the MAVISp repository directory, run:

streamlit run Welcome.py

a browser window displaying the MAVISp web app should open.

These instructions allow to run our web app on a minimal test dataset that is included in the distribution.

1. If you haven't already, activate your Python environment (see previous steps)

2. create a local copy of the MAVISp repository in your system:

git clone https://github.com/ELELAB/MAVISp

3. Create a copy of the test dataset in the MAVISp directory

cd MAVISp
rm -rf ./database
cp -r test_data/mavisp_web_server database

4. With your Python environment still active and from inside the MAVISp repository directory, run:

streamlit run Welcome.py

a browser window displaying the MAVISp web app should open.

the steps are the same as the last section, but a browser window will not open. Instead, you will have to connect from your local browser to the host printed out by streamlit in the terminal.

This option is useful for who wants to run MAVISp remotely, but doesn't have direct access to the MAVISp web service (e.g. because the outbound port that streamlit uses is blocked). It is however pretty slow and clunky, meaning it is not very effective for in-depth data exploration or analysis. It also requires to be able to connected with X forwarding to your server, and to have a web browser installed on your server. For a faster alternative that doesn't use X-forwarding see instructions below.

1. connect to your server via ssh, with X forwarding:

ssh -XY user@hostname

2. Follow the previous instructions to install requirements, download required files and run the app locally. A browser window should pop up.

You can use an ssh tunnel to connect to your streamlit instance. This is slightly more complicated but leads to an overall much better user experience.

1. ssh without X forwarding to the server:

ssh user@hostname

2. Follow the previous instructions to install the requirements and download the required files for MAVISp

3. run the app in headless mode:

cd MAVISp
streamlit run --logger.level=info --server.headless=true Welcome.py

Please note down the port that Streamlit is using (i.e. in this case it is 8501)

4. on your workstation open a new terminal and open an SSH tunnel:

ssh -N -L 8080:hostname:8501 user@hostname

notice that you need to change the port number at the right of hostname: with the one that Streamlit is providing service on

6. on your workstation open a new browser window and visit the website localhost:8080. MAVISp should load.

the MAVISp Python package is not necessary to run the web app - it is however necessary to generate the database file starting from raw input files. It includes a user-executable script (mavisp) with this very purpose, which performs sanity checks on the input data, prints a report, and performs the conversion. Please see the help text of the script itself for further details (mavisp -h) or instructions below.

The MAVISp Python package is designed to run on any operating system that supports Python. It has been tested on Ubuntu Linux 18.04 and macOS (13.5.2)

In order to install the package and all its requirements automatically, you will need to have a working Python 3.9+ installation available. We recommend installing the package in its own virtual environment - please see previous instructions on how to create a virtual environment.

The MAVISp Python package requires the following packages, and has been tested with the following versions:

• pandas 2.1.3
• tabulate 0.9.0
• matplotlib 3.7.4
• numpy 1.26.2
• PyYAML 6.0.1
• streamlit 1.28.2
• streamlit-aggrid 0.3.4.post3
• requests 2.31.0
• termcolor 2.3.0

Once your virtual environment is ready and active, you need to

1. create a local copy of the repository:

git clone https://github.com/ELELAB/MAVISp

2. install the Python package

cd MAVISp
pip install .

the mavisp executable will be available.

Installation time is typically up to a few minutes.

Notice that installing the Python package always installs all the requirements for the web app, meaning it is ready to run.

A typical command line of the mavisp script looks like:

mavisp -d input_data -o output_database

where input_data is a folder containing the raw data in a specific format and output_database is were the csv files database will be written. The script

1. performs all the parsing on the input file as well as basic sanity checks
2. prints a summary of all the available datasets and their status
3. prints a more detailed report of the status of each MAVISp module in each dataset
4. writes the output database. The output directories needs not to be present, unless option -f is set, which forces writing or overwriting the database.

To test the script on the MAVISp dataset included in the repository, from inside the MAVISp folder you have cloned from GitHub (see previous instructions), just run:

mavisp -d test_data/mavisp_python_package -o test_output

the test_output folder will now contain the output of the command. A reference output for this command can be found in the test_data/mavisp_web_server/ folder. The execution should complete in a few seconds.

The MAVISp modules can have 4 possible states:

• OK, colored in green, for when no warnings or errors were detected
• WARNING, colored in yellow, for when prcessing the module generated some messages that can be useful for the user, but was still able to read and elaborate the data for the module correctly
• ERROR, colored in red, for when processing the module resulted in a critical error i.e. a problem that couldn't be overcome
• NOT_AVAILABLE, colored in the default terminal color, when the module was not present in the dataset

Additionally, if the mutation list or metadata file are not available, the whole entry is flagged as being in a CRITICAL state - if this is the case, the corresponding modules will not be processed and the detected error will be displayed in the report.

If any module generates at least one error or warning for a given dataset, the status of the corresponding dataset in the summary table is set accordingly.

If any module generates an error, the script will not write the corresponding database file and exit.

It is possible to process only some of the available proteins in the dataset, or exclude some, by using options -p or -e respectively, with a comma-separated list, e.g.

mavisp -d input_data -o output_database -p MAP1LC3B,BCL2

it is also possible to perform a check of the raw data without writing the database files by using option -n.