This repository corresponds to the manuscript "Predictive Modeling of Microbial Data with Interaction Effects" (https://www.biorxiv.org/content/10.1101/2024.04.29.591596v1)
It contains code for quadratic interaction modeling across different microbial data types:
- Quantitative microbiome data (absolute counts)
- Presence-absence data
- Relative abundance data (compositional data)
Our framework unifies statistical approaches for microbial interactions, offering variable selection, hierarchical constraints, and stability-based model selection to ensure robust and interpretable models.
We demonstrate its versatility with applications to:
- Gut microbiome data (predicting antimicrobial resistance genes)
- Synthetic communities (inferring species-function relationships)
- Marine microbiomes (predicting environmental parameters)
The results in our manuscript can be reproduced using the provided RMarkdown files.
