MultiAddGPs (Multinomial Logistic Normal Additive Gaussian Process Model)

This repository contains code accompanying https://arxiv.org/abs/2410.03911

Brief Overview

MultiAddGPs is a powerful statistical model designed to analyze additive linear and nonlinear effects on compositional count data, particularly suited for relative abundance analysis in microbiome studies or similar high-dimensional compositional datasets. The model leverages the Multinomial Logistic-Normal distribution, combined with Additive Gaussian Processes, to capture complex mixed linear and non-linear dynamics.

Reproducibility

To reproduce the figures presented in the main text, follow the instructions below:

1. Clone the repository to your local machine:

git clone https://github.com/tinghua-chen/MultiAddGPs.git

2. Install the necessary R packages:

packages <- c("fido","dplyr","LaplaceDemon","mvtnorm","MASS","ggplot2","rBayesianOptimization","lubridate","gridExtra")
install.packages(packages)

3. Run 00_package_loading.R to load all necessary dataset
4. Run helper_functions.R to load all necessary functions

Reproducing Figures:

1. Run 01_data_preprocessing.R to load and preprocess dataset
2. Run 02_analysis.R to perform the required analysis on the dataset
3. Run 03_result.R to generate Figures.

Example

I have created a vignette demonstrating how to apply MultiAddGPs in practice using artificial gut data. You can view it here.

Note: MultiAddGPs extends the functionality of our previous fido package to accommodate both linear and nonlinear modeling in count data. If you haven't done so yet, I highly recommend reviewing the basset vignette in the fido package before working through the above example.

Contact

If you have any questions, don't hesitate to create an issue or reach out to tuc579@psu.edu