Abstract Emerging evidence suggests that human microbiome, composed of collective genomes of as many as 100 trillion microorganisms, could be mediating disease-leading causal pathways initiated by environmental toxicants or other factors such as drug usage. Prenatal arsenic exposure through drinking water, for example, could initiate perturbation of gut microbiome, and therefore, children could inherit perturbed microbiome composition if their mothers have arsenic exposure during perinatal period. The unhealthy microbiome composition could, in turn, induce children’s asthma, infection and allergy which could explain that arsenic exposure during pregnancy is related to children’s infection. Taken together, arsenic exposure could be the initiation of causal pathways leading to children’s infection through perturbed mother’s microbiome being passed to children. There are many other possible initiation factors such as diet, gene mutation, delivery mode and antibiotics leading to different children’s health outcomes. These mediations could happen through changes in particular microbial taxa or though the perturbation of microbiome population structure. While high-throughput sequencing technologies can characterize the taxonomic composition of microbiome in unprecedented detail, none of the existing mediation analysis methods is adequate enough to model the mediation effects of microbiome due to the unique challenging features of microbiome data. Therefore, there is an urgent need to have appropriate mediation analysis methods in place for estimating and testing the mediational effects of human microbiome. To address these issues, we will develop two general mediation analysis frameworks to identify mediation through changes in individual microbial taxa and model mediation though the perturbation of overall microbiome composition. The models will be tested with extensive simulations and cross validations. An R package and an interactive web application will be developed for model implementations. In the real study applications, we will quantify and test the mediation effects of infant gut microbiome and breast-milk microbiome in the relations between prenatal exposures (e.g., arsenic exposure, maternal diet) and childhood infections and allerg/atopy in the first year of life using the rich data from the large ongoing longitudinal molecular epidemiologic New Hampshire Birth Cohort study. With the applications of the proposed models in a cystic fibrosis (CF) study, we will examine whether CF transmembrane conductance regulator gene mutations lay the biological foundation for patterns in the developing microbiome in the gut that are associated with CF exacerbation onset in newborn children with CF. By analyzing the data from the Infant Growth and Microbiome Study, we will answer the key question...