TABLE 1.

Characteristic and zero-value-treatment methods of the commonly used microbial association network construction methods.

Method	Characteristic	Zero value treatment
SparCC (6)	SparCC focuses on correlation inference of compositional data, and uses Aitchison’s variance of log-ratios to solve ingredient problems. It could be used on any genomic survey data that has low diversity.	Add a pseudo-value of 1 to each element.
MENA (7)	MENA uses Random Matrix Theory (RMT) to identify a cutoff for constructing microbial association networks, by determining the transition point of nearest-neighbor spacing distribution of eigenvalues from Gaussian (random) to Poisson (non-random) distribution.	By default, unpaired zero values were filled by a pseudo-value of 0.01. More user defined options are also available.
LSA (8)	LSA breaks down the global molecular similarity as local similarity at each grid point surrounding the molecules and is efficient to calculate the statistical significance for pairwise local similarity analysis, making possible all-to-all local association studies otherwise prohibitive. It is commonly used to construct association network from time series data.	No mention of the treatment of zero values.
CoNet (9)	CoNet offers ensemble-based network construction, by combining a number of different correlations (Pearson, Spearman, and Kendall), similarities (Mutual information) or dissimilarities (Bray-Curtis and Kullback-Leibler) to improve the accuracy of the strength of the associations between objects.	Omitting sample pairs with zero values from the association strength calculation.
SPIEC-EASI (10)	SPIEC-EASI is a computational framework that includes statistical methods for the inference of microbial ecological interactions from 16S rRNA gene sequencing datasets. A sophisticated synthetic microbiome data generator with controllable underlying species interaction topology is also equipped.	Add a pseudo-value of 1 to zero values.