Sparse Estimation of Correlations among Microbiomes (SECOM)

Abstract It is well-known that human gut microbiota form a complex ecosystem where microbes correlate with each other. Due to complex underlying interactions, some microbes may correlate nonlinearly. Yet, there are no measures in the literature that quantify these nonlinear relationships. We introduce a novel concept of correlation that allows for linear and nonlinear relationships among microbes. It is well-known that infant gut evolves as the child grows. Using the proposed methodology, for the first time in the literature we characterize temporal changes in correlations among bacterial families during a baby’s first year after birth. For example, we discover for the first time that Lachnospiraceae and Ruminococcaceae, which play important roles in degrading complex polysaccharides into short-chain fatty acids, interact with several families during a baby’s first year of birth. Similarly, Bifidobacteriaceae, Bacteroidiceae, Lachnospiraceae, and Clostridiaceae, are all negatively correlated with Enterobacteriaceae, a family that contains pathogens such as Salmonella, Shigella and Escherichia coli. Statistical properties of the proposed methodology are evaluated analytically and by using synthetic data.