Accounting for missing actors in interaction network inference from abundance data

TL;DR

This paper introduces a novel mixture model with a variational EM algorithm to infer dependency structures in count data, accounting for missing actors, and demonstrates its effectiveness on ecological datasets.

Contribution

It presents a new mixture of Poisson log-normal distributions with tree-shaped graphical models to recover missing actors in network inference from abundance data.

Findings

Successfully recovers environmental drivers in ecological datasets.

Performs well on synthetic data for dependency structure recovery.

Provides an R package for implementation.

Abstract

Network inference aims at unraveling the dependency structure relating jointly observed variables. Graphical models provide a general framework to distinguish between marginal and conditional dependency. Unobserved variables (missing actors) may induce apparent conditional dependencies.In the context of count data, we introduce a mixture of Poisson log-normal distributions with tree-shaped graphical models, to recover the dependency structure, including missing actors. We design a variational EM algorithm and assess its performance on synthetic data. We demonstrate the ability of our approach to recover environmental drivers on two ecological datasets. The corresponding R package is available from github.com/Rmomal/nestor.