Abstract:In multivariate statistics, the question of finding direct interactions can be formulated as a problem of network inference - or network reconstruction - for which the Gaussian graphical model (GGM) provides a canonical framework. Unfortunately, the Gaussian assumption does not apply to count data which are encountered in domains such as genomics, social sciences or ecology. To circumvent this limitation, state-of-the-art approaches use two-step strategies that first transform counts to pseudo Gaussian observations and then apply a (partial) correlation-based approach from the abundant literature of GGM inference. We adopt a different stance by relying on a latent model where we directly model counts by means of Poisson distributions that are conditional to latent (hidden) Gaussian correlated variables. In this multivariate Poisson lognormal-model, the dependency structure is completely captured by the latent layer. This parametric model enables to account for the effects of covariates on the counts. To perform network inference, we add a sparsity inducing constraint on the inverse covariance matrix of the latent Gaussian vector. Unlike the usual Gaussian setting, the penalized likelihood is generally not tractable, and we resort instead to a variational approach for approximate likelihood maximization. The corresponding optimization problem is solved by alternating a gradient ascent on the variational parameters and a graphical-Lasso step on the covariance matrix. We show that our approach is highly competitive with the existing methods on simulation inspired from microbiological data. We then illustrate on three various data sets how accounting for sampling efforts via offsets and integrating external covariates (which is mostly never done in the existing literature) drastically changes the topology of the inferred network.

Estimating graphical models for count data with applications to single-cell gene network

Gene regulatory network in single cells based on the Poisson log-normal model

Network Analysis of Count Data from Mixed Populations

Simultaneous Estimation of Many Sparse Networks via Hierarchical Poisson Log-Normal Model

A Log-Linear Graphical Model for Inferring Genetic Networks from High-Throughput Sequencing Data

Single-cell gene regulatory network analysis for mixed cell populations with applications to COVID-19 single cell data

L1-Penalized Censored Gaussian Graphical Model

Model selection and estimation in the matrix normal graphical model

Sparse Inverse Covariance Estimation for High-throughput microRNA Sequencing Data in the Poisson Log-Normal Graphical Model

Generalized Precision Matrix for Scalable Estimation of Nonparametric Markov Networks

Covariate-adjusted Precision Matrix Estimation with an Application in Genetical Genomics

Variational inference for sparse network reconstruction from count data

Using Matrix of Thresholding Partial Correlation Coefficients to Infer Regulatory Network

GMFGRN: a matrix factorization and graph neural network approach for gene regulatory network inference

Simultaneous clustering and estimation of networks in multiple graphical models

Network analysis for count data with excess zeros

PLNseq: a multivariate Poisson lognormal distribution for high-throughput matched RNA-sequencing read count data.

A sparse conditional Gaussian graphical model for analysis of genetical genomics data

Dissecting and improving gene regulatory network inference using single-cell transcriptome data

A sparse negative binomial mixture model for clustering RNA-seq count data

Approximate inference of gene regulatory network models from RNA-Seq time series data