MDITRE: scalable and interpretable machine learning for predicting host status from temporal microbiome dynamics

We present a new differentiable model that learns human interpretable rules from microbiome time-series data for classifying the status of the human host.

Installation (Python 3.6+ and CUDA 10.0)

First install mditre package from pip via web or source. Then install pytorch.

Install mditre from pip via web:

pip install mditre

Install mditre from pip via source:

git clone https://github.com/gerberlab/mditre.git
cd mditre
pip install .

Install pytorch from pip

Linux or Windows (with NVIDIA GPU and CUDA 10.0)

pip install torch==1.4.0 torchvision==0.5.0 -f https://download.pytorch.org/whl/cu100/torch_stable.html

Linux or Windows (CPU only)

pip install torch==1.4.0+cpu torchvision==0.5.0+cpu -f https://download.pytorch.org/whl/torch_stable.html

MacOS (CUDA not supported)

pip install torch==1.4.0 torchvision==0.5.0

Usage

MDITRE workflow on 16S rRNA and shotgun metagenomics data

We provide 2 tutorials, one for 16s-based data here and another for shotgun metagenomics (Metaphlan) based data here, which show how to use MDITRE for data loading and preprocessing, running the model code and using the GUI to interpret the learned rules for post-hoc analysis.

Configuration options

MDITRE operation requires a list of configuration options to be passed as arguments as explained here.