Su, C.; He, M.; Van Doren, V. E.; Kelley, C. F.; Hu, Y.

Most existing methods for inferring microbial networks generate only point estimates of Pearson\'s correlations without assessing their significance, and none can account for clustered samples. In this article, we introduce TestNet, a novel method that delivers well calibrated results by controlling the false discovery rate (FDR). We developed a permutation-based procedure to generate valid null replicates that account for compositional effects and extensive zeros in microbiome data, and clustering structures within the samples when present. Our results demonstrate that TestNet is the only method that effectively controls the FDR while maintaining high power across a wide range of scenarios.