Network models such as graphical models have become a useful approach to studying the interactions between microbial taxa given the microbiome data deluge. Recently, various methods for sparse inverse covariance estimation have been proposed to estimate graphical models in the high-dimensional setting, including graphical lasso. However, current methods do not address the compositional count nature of microbiome data, where abundances of microbial taxa are not directly measured but are presented by error-prone counts. Adding to the challenge is that the sum of the counts within each sample, termed “sequencing depth”, can vary drastically across samples. To address these issues, we adopt a logistic normal multinomial model explicitly incorporating the sequencing depth and develop an algorithm iterated between Newton-Raphson and graphical lasso for model estimation. We call this new approach “compositional graphical lasso”. We have established the convergence of the algorithm. Additionally, we illustrate the advantage of compositional graphical lasso in comparison to current methods under a variety of simulation scenarios and also demonstrate the applicability of compositional graphical lasso to a human microbiome data set.