Microbiota forms complex community structures and collectively affect human health. Understanding the structural organization of the human microbiome plays a vital role in revealing how the microbial taxa are collaborating or competing with each other under different host physiology conditions. The exponentially growing large datasets made available by next-generation sequencing technology motivate the development of statistic tools to quantitatively study the microbial community structures. To this end, we proposed a general framework, HARMONIES, a Hybrid Approach foR MicrobiOme Network Inferences via Exploiting Sparsity, to infer the microbial association networks.

HARMONIES accounts for the uneven sequencing depth, zero-inflation, and over-dispersion when normalizing the real microbiome data. It also provides a sparse and consistent estimation of the network structure by selecting the most stable associations between taxa in the network.