Causal inference and discovery for the microbiome (CARDAMOM)

A wealth of recent studies has led to a widespread optimism that the microbiome holds the key to both deciphering complex inflammatory and immunological diseases, as well as finding cures and preventive measures. Despite of all the consequent excitement, there is still a dearth of verified discoveries about the mechanistic role of named microbes and their specific gene products in the complex disease etiology. CARDAMOM will enable causal discovery and inference for the microbiome by developing novel statistical tools relying on advanced modeling and computation, combined with biological and epidemiological expertise. In one project arm we will generalize existing families of graphical models to meet the needs of causal analysis for microbiome and rigorously study their statistical properties. In the other arm we will develop methods for high-dimensional likelihood-free inference to enable predictions and causal quantification from dynamical microbiome system models based on data. The new tools will be disseminated via the ELFI software platform (www.elfi.ai), which is an internationally leading open-source software project on enabling inference for simulator-based models. CARDAMOM will also generate pediatric nasopharyngeal and gut microbiome cohort data and apply the developed tools to significantly advance our understanding about the microbial ecosystem dynamics and the consequences of interventions. We anticipate that our new tools will be widely useful for the general microbiome research community, and for a large body of researchers across multiple fields of engineering, medicine and science.

A wealth of recent studies has led to a widespread optimism that the microbiome holds the key to both deciphering complex inflammatory and immunological diseases, as well as finding cures and preventive measures. Despite of all the consequent excitement, there is still a dearth of verified discoveries about the mechanistic role of named microbes and their specific gene products in the complex disease etiology. CARDAMOM will enable causal discovery and inference for the microbiome by developing novel statistical tools relying on advanced modeling and computation, combined with biological and epidemiological expertise. In one project arm we will generalize existing families of graphical models to meet the needs of causal analysis for microbiome and rigorously study their statistical properties. In the other arm we will develop methods for high-dimensional likelihood-free inference to enable predictions and causal quantification from dynamical microbiome system models based on data. The new tools will be disseminated via the ELFI software platform (www.elfi.ai), which is an internationally leading open-source software project on enabling inference for simulator-based models. CARDAMOM will also generate pediatric nasopharyngeal and gut microbiome cohort data and apply the developed tools to significantly advance our understanding about the microbial ecosystem dynamics and the consequences of interventions. We anticipate that our new tools will be widely useful for the general microbiome research community, and for a large body of researchers across multiple fields of engineering, medicine and science.