HOMINID

Python MPI program to identify associations between host genetic variation and microbiome taxonomic composition.

Summary: Recent studies have uncovered a strong effect of host genetic variation on the composition of host-associated microbiota. Here, we present HOMINID, a computational approach based on Lasso linear regression, that given host genetic variation and microbiome composition data, identifies host SNPs that are correlated with microbial taxa abundances. By using HOMINID on data from the Human Microbiome Project, we identified 13 human SNPs in which genetic variation is correlated with microbiome taxonomic composition in 15 body sites.

We also present a tool for visualization of host-microbiome association network identified in HOMINID, currently including toy data representing all SNP-microbe associations with a nominal p-value <= 0.1. Online visualization tool at http://z.umn.edu/genemicrobe

Contact: [email protected]

Running the HOMINID software

This README describes the installation process and how to test the HOMINID software on included test data. Once HOMINID is installed and known to work read the following documents for instructions on using it with your own data:

1. HOMINID analysis pipeline
2. Running hominid on your data
3. Running hominid_stability_selection on your data
4. Running hominid_sort_results on your data

Requirements

HOMINID is a Python 3.6+ MPI program. It is intended to run on a cluster, but it will run anywhere with a working MPI implementation and mpi4py. HOMINID has been tested only on Linux operating systems.

The required Python packages will be automatically installed. They are:

• mpi4py (version 2.0 or greater)
• numpy
• pandas
• scipy
• scikit-learn (version 0.19.1)
• scikits.bootstrap

The optional plotting script requires R 3.2+ and rpy2. These can be installed using the Anaconda Python distribution with the r-essentials packages as shown below.

Computer requirements

HOMINID is a multiprocess program that benefits from multiple cores and multiple processors. It can run on any hardware that supports mpi4py from laptops to clusters.

Install

It is recommended that HOMINID be installed in a Python virtual environment. These instructions are specifically for the Miniconda3 distribution, which has been tested with HOMINID.

Install a MPI Implementation

A MPI implementation is available on most clusters so this step is generally necessary only on laptop and desktop computers. HOMINID is known to run with OpenMPI on Ubuntu 14.04, 16.04, and 18.04. These commands will install OpenMPI on Ubuntu and similar Debian-based Linux distributions:

$ sudo apt update
$ sudo apt install mpi-default-dev openmpi-bin

Create a Python Virtual Environment

1. Download and install Miniconda3 from a terminal with these commands:

$ wget http://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh -O miniconda3.sh
$ chmod u+x miniconda3.sh
$ ./miniconda3.sh

• You will be asked to accept the license as part of the installation process.
• Accept the default installation directory unless you have good reason not to. The default directory is ~/miniconda3.
• The installer will ask to modify your PATH variable to include conda. Choose yes. If you choose no you will need to use full pathnames in subsequent steps.

2. Close the terminal and open a new one so the update to your PATH variable takes effect. In the new terminal update conda:

$ conda update conda

3. Clone the hominid repository.

$ git clone https://github.com/blekhmanlab/hominid.git
$ cd hominid

4. Create a new virtual environment and install the HOMINID software. Here the virtual environment is named hom but another name will work.

$ conda create -n hom python=3.6 --file conda-requirements.txt
$ conda activate hom
(hom) $ pip install .
(hom) $ conda install rpy2 r-essentials

The pip install command installs the HOMINID package itself and a package that is not available to conda. The final conda install command installs packages for the optional plotting script.

Once HOMINID has been installed with pip the scripts can be executed from any directory by name as follows:

(hom)$ hominid
(hom)$ hominid_stability_selection
(hom)$ hominid_sort_results
(hom)$ hominid_box_bar_plot

The Python programs are in directory hominid/hominid. Test scripts are in directory hominid/example/scripts. Test input data files are in directory hominid/example/data.

Test the Installation

The installation can be tested using the included test scripts and test data with the following steps.

1. Run hominid on the sample data. Change directory to the example/scripts directory and run test_hominid.sh.

(hom)$ cd example/scripts
(hom)$ ./test_hominid.sh

In test_hominid.sh, the option -n 3 to mpirun specifies that 3 processes will be used. Change this if you want to use a different number of processes. Performance will be reduced if more processes are specified than available cores. A minimum of two processes must be specified. The test output is written to hominid/example/hominid_example_output.rvcf. Many lines will be printed to stderr so you can watch hominid's progress.

2. Run hominid on the sample data, with permuted sample IDs:

(hom)$ ./test_hominid_permute.sh

Output is written to hominid/example/hominid_example_output_permute.rvcf.

3. Run stability selection to find associated OTUs/taxa/covariates.

(hom)$ ./test_stability_selection.sh

Output is written to hominid/example/stability_selection_example_output.rvcf.

4. Combine the Lasso regression results with the microbiome abundances:

(hom)$ ./test_sort_results.sh

Output is written to hominid/example/sort_results_example_output.