SIGNATURE WORK
CONFERENCE & EXHIBITION 2023

This study aims to establish a novel computational framework for determining causal relationships between human gut microbial species and metabolic diseases, with a focus on Type 2 Diabetes. The research offers a comprehensive review of existing correlation studies, emphasizing the distinction between causal inference and correlation analyses. By applying the Potential Outcomes Framework, we develop an innovative, quantitatively causal pipeline that relies solely on observational data, specifically relative abundance. The Guangzhou Nutrition Health Study serves as the main cohort, while the Zhejiang Metabolic Syndrome Cohort acts as a validation cohort. Models such as Meta Learner, Doubly Robust Learner, and Uplift Model are employed to estimate Individual Treatment Effects. Average Treatment Effect is utilized for ranking purposes. Our findings reveal the top 50 microbial species with strong causal associations to Type 2 Diabetes, demonstrating that over 50% of these species exhibit both causal and correlated connections to the disease. Further downstream analyses of disease outcomes and medication support the relationships between top species and diabetes. A literature review confirms that our computational results align with the general roles of most prominent species in the human gut. Additionally, we identify 10 cross-cohort species that play consistent roles in both our computational findings and previous research. The significance of this study lies in the development of a new analysis pipeline that employs a causal model to assess the gut microbiome in metabolic-driven human disease phenotypes, laying the groundwork for future research in this area.