MACH2

MACH2 is a computational tool for inferring optimal migration histories of metastatic cancer from clonal trees and observed clone locations. It supports multiple parsimony criteria, including the unobserved clone criterion, and guarantees optimality and completeness by systematically enumerating solutions.

Preprint: bioRxiv.

For the accompanying visualization tool, see MACH2-viz.

Table of contents

1. Installation

    1.1 [Prerequisite](#11-prerequisite)  
    1.2 [Install using `pip`](#12-install-using-pip)  
    1.3 [Install using `conda`](#13-install-using-conda)  
    1.4 [Install from source](#14-install-from-source)  

2. Usage instruction

    2.1 [I/O formats](#21-io-formats)  
    2.2 [Usage](#22-usage)  
            2.2.1 [From `JupyterLab`](#221-from-jupyterlab)  
            2.2.2 [From Terminal](#222-from-terminal)  

3. Reproduction

1. Installation

1.1 Prerequisites

• Python - MACH2 requires Python 3.12 .
• ILP solver - MACH2 requires an ILP solver installed. Currently MACH2 only supports Gurobi optimizer (version 12 or newer), but we are going to add support for more ILP solvers in the future. MACH2 requires a valid Gurobi installation and license key. The location of Gurobi should be present in LD_LIBRARY_PATH (linux) or DYLD_LIBRARY_PATH (macOS), and the license key should be saved in the environment variable GRB_LICENSE_KEY.
• OS - Since MACH2 is a pure Python package, it can be run on any operating system that supports Python.

MACH2 has been tested with Python 3.12 and Gurobi optimizer version 12 on CentOS 7.

1.2 Install using pip

MACH2 can easily be installed using pip, the package installer for Python. Gurobi python package also needs to be installed separately. Open a terminal or command prompt and run the following command:

            $ pip install gurobipy
            $ pip install mach2

If you want to install MACH2-viz along with MACH2, you can run the following commands instead:

            $ pip install gurobipy
            $ pip install 'mach2[viz]' 

MACH2-viz can be installed separately too.

1.3 Install using conda

MACH2 can be installed using conda from the channel bioconda. We advise to create a new environment in conda.

            $ conda create -n mach2 python=3.12 gurobi mach2 -c gurobi -c bioconda

If using existing conda environment, the following command installs MACH2 and Gurobi.

            $ conda install gurobi mach2 -c gurobi -c bioconda

1.4 Install from source

Since MACH2 is a python package, one may clone the repository and install from source using pip.

            $ git clone https://github.com/elkebir-group/MACH2.git
            $ cd MACH2
            $ pip install .

In any approach, MACH2 takes less than 5 minutes to install on a normal computer with regular internet speed.

2. Usage Instruction

2.1 I/O formats

We describe various formats used by MACH2.

1. Tree file : The tree file contains a list of edges that define the structure of a tree. Each line in the file represents an edge, and the edges should be in the format: node1 node2. For example:

    1   2
    2   3
    2   4 
    3   5

2. Tree file with timing/comigrations : Tree file with timestamps. Edges with the same timestamp belong to the same comigration, and a timestamp with -1 represents non-migration. Each line corresponds to an edge in the format: node1 node2 timestamp. For example:

    1   2   -1
    2   3   1
    2   4   1
    3   5   2

3. Observed labeling file : The observed labeling file contains zero or more location labels assigned to each node of the input clonal tree. Each line in the file corresponds to a node and the labels assigned to it in the format: node label1 label2 .... If a node is not observed anywhere, it may be skipped. For example:

    1   A   B
    3   B
    4   A   C
    5   C

4. Location labeling file : The location labeling file contains the unique location label of origin assigned to each node. Each line in the file corresponds to a node and the location label of origin are in the format: node label. For example:

    1   A
    2   B
    3   C

5. Node of origin file : The node of origin file maps the nodes of the refined tree to the nodes of the input clonal tree. Each line in the file corresponds to a vertex and the labels are in the format: leaf label. For example:

    1   A
    2   B
    3   C

Additionaly, MACH2 can output files in Graphviz DOT format or JSON format.

2.2 Usage

MACH2 takes as input two files -

1. Tree file : Tree file describing the input clone tree.
2. Observed labeling file : Labeling file describing the observed labeling of input clone tree.

MACH2 Can be run using command line, or can be directly accessed from JupyterLab.

2.2.1 From JupyterLab

MACH2 can be directly imported and accessed from JupyterLab. We include an [example JupyterLab notebook][https://github.com/elkebir-group/MACH2/blob/main/analysis/example.ipynb] to illustrate basic usage. The notebook runs in less than one minute on a normal computer. For more details, check the documentation for classes and methods.

2.2.2 From Terminal

For each solution, MACH2 can output three types of files.

1. Tree file with timing/comigrations : Refined tree file with timestamps/comigrations.
2. Location labeling file : Location labeling file describing the location labeling of the refined tree.
3. Node of origin file : Node of origin file mapping refined tree nodes to input clonal tree nodes.

Additionaly MACH2 can return JSON file encoding all the solutions. The JSON file can be directly passed to MACH2-viz. The exact format of the JSON file is described here.

MACH2 also prints <primary location> <number of migrations> <number of comigrations> Optimal <running time (in seconds)> on console.

MACH2 can be run using python.

            mach2 [-h] [-c CRITERIA] [-p PRIMARY] [--colormap COLORMAP] [--log] [-o OUTPUT] [--starting_nsols STARTING_NSOLS] [--max_solutions MAX_SOLUTIONS] [-t THREADS]
                    [--timelimit TIMELIMIT] [--viz]
                    clonal_tree observed_labeling

            MACH2

            positional arguments:
            clonal_tree           Input clonal tree
            observed_labeling     Input observed labeling

            optional arguments:
            -h, --help            show this help message and exit
            -c CRITERIA, --criteria CRITERIA
                                    Criteria ordering
            -p PRIMARY, --primary PRIMARY
                                    Primary anatomical location
            --colormap COLORMAP   Color map file
            --log                 Outputs Gurobi logging (default=False)
            -o OUTPUT, --output OUTPUT
                                    Output folder (default=current folder)
            --starting_nsols STARTING_NSOLS
                                    Starting number of solutions (default=37)
            --max_solutions MAX_SOLUTIONS
                                    Maximum number of solutions retained (default=37888)
            -t THREADS, --threads THREADS
                                    Number of threads
            --timelimit TIMELIMIT
                                    Returns the best suboptimal solution found in the time limit (in seconds).
            --viz, --open_in_viz  Open the locations on MACH2-viz (default=False)

An example execution

    $ mach2 data/breast/A1.tree data/breast/A1.observed.labeling -p breast --colormap data/breast/coloring.txt
    

For this specific instance, MACH2 takes less than one second to run on a normal computer.

3. Reproduction

We include a JupyterLab notebook containing relevant codes to run MACH2 (and also the other migration history inference methods) on all the simulated and real data and generate the results presented in the preprint.