Sparse and compositionally robust inference of microbial ecological networks Zachary D. Kurtz, Christian L. Mueller, Emily R. Miraldi, Dan R. Littman, Martin J. Blaser, and Richard A. Bonneau present SPIEC-EASI, a statistical method for inferring microbial ecological networks from amplicon sequencing data. The method addresses two key challenges: compositional data biases and the underdetermined data regime (p > n). SPIEC-EASI combines data transformations from compositional data analysis with a graphical model inference framework that assumes the underlying network is sparse. It uses algorithms for sparse neighborhood and inverse covariance selection to reconstruct the network. SPIEC-EASI is accompanied by computational tools to generate synthetic OTU data from diverse network topologies. It outperforms state-of-the-art methods in recovering edges and network properties on synthetic data and reproducibly predicts previously unknown microbial associations using data from the American Gut project. The study highlights the challenges of analyzing microbial community data, including compositional biases and the underdetermined data regime. SPIEC-EASI uses conditional independence to infer an underlying graphical model, avoiding detection of indirect network interactions. It is compared to other methods like SparCC and CCREPE, which rely on correlation analysis. SPIEC-EASI's performance is evaluated on synthetic data generated from diverse network topologies, including band-like, cluster, and scale-free graphs. The results show that SPIEC-EASI outperforms other methods in recovering network structures, particularly in high-dimensional scenarios. It also demonstrates strong performance in recovering microbial associations from real data, including clusters of strongly connected OTUs and the presence of archetypical network topologies in gut microbial networks. SPIEC-EASI is implemented in the R package SpiecEasi and is available for use in microbiome studies.Sparse and compositionally robust inference of microbial ecological networks Zachary D. Kurtz, Christian L. Mueller, Emily R. Miraldi, Dan R. Littman, Martin J. Blaser, and Richard A. Bonneau present SPIEC-EASI, a statistical method for inferring microbial ecological networks from amplicon sequencing data. The method addresses two key challenges: compositional data biases and the underdetermined data regime (p > n). SPIEC-EASI combines data transformations from compositional data analysis with a graphical model inference framework that assumes the underlying network is sparse. It uses algorithms for sparse neighborhood and inverse covariance selection to reconstruct the network. SPIEC-EASI is accompanied by computational tools to generate synthetic OTU data from diverse network topologies. It outperforms state-of-the-art methods in recovering edges and network properties on synthetic data and reproducibly predicts previously unknown microbial associations using data from the American Gut project. The study highlights the challenges of analyzing microbial community data, including compositional biases and the underdetermined data regime. SPIEC-EASI uses conditional independence to infer an underlying graphical model, avoiding detection of indirect network interactions. It is compared to other methods like SparCC and CCREPE, which rely on correlation analysis. SPIEC-EASI's performance is evaluated on synthetic data generated from diverse network topologies, including band-like, cluster, and scale-free graphs. The results show that SPIEC-EASI outperforms other methods in recovering network structures, particularly in high-dimensional scenarios. It also demonstrates strong performance in recovering microbial associations from real data, including clusters of strongly connected OTUs and the presence of archetypical network topologies in gut microbial networks. SPIEC-EASI is implemented in the R package SpiecEasi and is available for use in microbiome studies.