1. Getting started#

This page explains how to install Maud and use its command line interface.

1.1. Installing Maud#

Maud is compatible with Python versions 3.9 and above

We recommend using a fresh virtual environment to install Maud. To make one and then activate it, run the following commands:

python -m venv .venv --prompt=maud
source .venv/bin/activate

To install the latest version of Maud and its python dependencies to your new virtual environment, run this command:

pip install maud-metabolic-models

Now Maud is installed!

1.1.1. Installing cmdstan#

Maud depends on the C++ application cmdstan, which in turn requires a c++ toolchain.

To save users the hassle of installing one of those, we ship Maud with pre-built binary files that make it just work on many computers. To see if your computer is one of these, try running this command after installing Maud:

maud simulate

If some output appears, then you don’t need to do anything!

If you see an error message, you will need to install cmdstan in order to use Maud. See here for a guide to how to do this using the Python package cmdstanpy, which should be installed after running pip install maud-metabolic-models.

1.2. Using Maud#

Maud is intended to be used by running commands starting with maud from the command line.

To see the available commands, type

maud --help

Maud’s main commands take a path to an input folder as their only argument. See the inputting page for how to create input folders.

Maud provides a few pre-made inputs and the command maud load-input <name> for loading them. You can see the available inputs by running maud load-input --help. Try loading the linear input as with this command:

maud load-input linear

A typical workflow for using Maud, after creating an input folder, begins by running the command maud simulate path_to_my_input. This will run Maud’s statistical model in simulation mode, generating a single snapshot of the metabolic steady state corresponding to the initial parameter values. This is useful for checking that the model and initial parameter values are sensible before starting a sampling run.

Try doing this with the default input by running this command:

maud simulate linear

Maud should display some interesting information to your console. It should also create a folder whose name starts with maud_output_sim_linear. This contains all the information you might need for downstream analysis.

The next step is to run the command maud sample to trigger a sampling run. You can do this as follows:

maud sample linear

When the sampler finishes, another folder should appear with a name beginning maud_output_linear.

To make out of sample predictions, given a sampling run you can use the command maud predict as follows, replacing <your-timestamp> with the actual timestamp:

maud predict maud_output_linear<your-timestamp>

When this finishes the output folder should be augmented with prediction information.