Compositional zero-inflated network estimation for microbiome data

Abstract

Background The estimation of microbial networks can provide important insight into the ecological relationships among the organisms that comprise the microbiome. However, there are a number of critical statistical challenges in the inference of such networks from high-throughput data. Since the abundances in each sample are constrained to have a fixed sum and there is incomplete overlap in microbial populations across subjects, the data are both compositional and zero-inflated. Results We propose the COmpositional Zero-Inflated Network Estimation (COZINE) method for inference of microbial networks which addresses these critical aspects of the data while maintaining computational scalability. COZINE relies on the multivariate Hurdle model to infer a sparse set of conditional dependencies which reflect not only relationships among the continuous values, but also among binary indicators of presence or absence and between the binary and continuous representations of the data. Our simulation results show that the proposed method is better able to capture various types of microbial relationships than existing approaches. We demonstrate the utility of the method with an application to understanding the oral microbiome network in a cohort of leukemic patients. Conclusions Our proposed method addresses important challenges in microbiome network estimation, and can be effectively applied to discover various types of dependence relationships in microbial communities. The procedure we have developed, which we refer to as COZINE, is available online at https://github.com/MinJinHa/COZINE.

Dysbiotic microbiome variation in colorectal cancer patients is linked to lifestyles and metabolic diseases

“…The interaction of microbiome relative abundance was visualized using a network analysis approach. Given the compositional and zero-inflated properties of the microbiome data, numerous correlation-focused approaches have been developed to overcome the difficulty of inferring dependencies in microbial data, such as CCREPE, SparCC, CCLasso, and REBACCA [ 21 , 22 ]. However, these methods may be limited in reflecting indirect relationships and causing spurious associations from the creation of pseudo-counts [ 21 , 22 ].…”

“…Given the compositional and zero-inflated properties of the microbiome data, numerous correlation-focused approaches have been developed to overcome the difficulty of inferring dependencies in microbial data, such as CCREPE, SparCC, CCLasso, and REBACCA [ 21 , 22 ]. However, these methods may be limited in reflecting indirect relationships and causing spurious associations from the creation of pseudo-counts [ 21 , 22 ]. Ha et al introduced a COpositional Zero-Inflated Network Estimation (COZINE) to address this challenge by generating a binary incidence matrix and a compositional abundance matrix in which the centered log-ratio transformation can be applied for non-zero counts only [ 21 , 22 ].…”

“…However, these methods may be limited in reflecting indirect relationships and causing spurious associations from the creation of pseudo-counts [ 21 , 22 ]. Ha et al introduced a COpositional Zero-Inflated Network Estimation (COZINE) to address this challenge by generating a binary incidence matrix and a compositional abundance matrix in which the centered log-ratio transformation can be applied for non-zero counts only [ 21 , 22 ]. In this study, the network structure was constructed for each group of smoking status, alcohol consumption, BMI, and underlying diseases using the COZINE method (R packages ‘COZINE’ and ‘HurdleNormal’) [ 21 , 22 ].…”

Risk of bias assessment tool for systematic review and meta-analysis of the gut microbiome

“…Interestingly, relatively few studies utilise any statistical modelling to correct for such missing data. Yet, various modelling techniques were recently developed to address zero-inflation (Ha et al, 2020; Zhang et al, 2020; Deek and Li, 2021). Similar to modelling techniques, imputation is a method traditionally used to address missing data in the form of patient drop out, but a promising imputation method is recently available to also deal with GMB sampling zeroes.…”

The MTIST platform: a microbiome time series inference standardized test simulation, dataset, and scoring systems

“…Each method parameterizes the generalized Lotka-Volterra (gLV) equations, thereby uncovering which directional pairwise interaction (e.g., +/+, -/-, +/-, 0/+, and 0/-) underlies each species pair in a microbial community ( Figure 1 A ). Network-based methods, on the other hand, infer non-directional associations between species without insight on the ecological niche between the species (Kurtz et al 2015; Nagpal et al 2020; Dhariwal et al 2017; Friedman and Alm 2012; Ha et al 2020; Fang et al 2015; Faust et al 2012). While bacterial associations are important to answer certain scientific questions, uncovering directional interactions is imperative to engineer precision microbiota-focused therapies.…”