Transkingdom Network Analysis (TkNA): a systems approach inferring causal factors underlying host-microbiota interactions

Technological advances have generated tremendous amounts of high-throughput omics data. Integrating data from multiple cohorts and/or several omics types from new and previously published studies can offer a holistic view of a biological system and decipher its critical players and key mechanisms. In this protocol we describe how to use TkNA, a unique causal-inference analytical framework that can perform meta-analysis of cohorts and detect master regulators (causal factors) among measured parameters (e.g., microbes or microbial genes, host genes, metabolites) that govern pathological or physiological responses of host-microbiota interactions in a particular condition or disease. The first step reconstructs the network that represents a statistical model capturing the complex relationships between the different omics of the biological system. Here, it selects differential features and their per-group correlations by identifying robust and reproducible patterns of fold change direction and sign of correlation across several cohorts. Next, using a causality-reflecting metric (correlation inequalities), statistical thresholds (per cohort p-values, combined Fisher’s p-value, and BH FDR), and a set of topological criteria (e.g., network sparsity, ratio of negative to positive correlations, and strengths of correlations) it selects the final edges that form the Transkingdom Network. The second part of the analysis involves interrogating the network. Using the network’s local (degree) and global (Bipartite Betweenness Centrality (BiBC)) topology metrics it detects nodes that are responsible for control of given subnetwork or control of communication between kingdoms and/or subnetworks. The underlying basis of TkNA approach involves fundamental principles including laws of causality, graph theory and information theory. Hence, TkNA can be used for causal inference via network analysis of any host and/or microbiota multi-omics data. This protocol takes approximately few hours to days to run, depending on size of available data and computational resources. It is targeted towards researchers that might not have access to high-performance computing and requires very basic familiarity with the Unix command-line environment. ### Competing Interest Statement The authors have declared no competing interest.