C-BIRD: Causal Biostatistical Framework for Interpretable and Robust Discovery of Polymicrobial Disease Biomarkers in the Human Microbiome

• Applicant: Dr. Sven Kleine Bardenhorst
• Subject Area: Epidemiology and Medical Biometry/Statistics
• Term: since 2026
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 576733886

The microbiome is a key factor in health and disease, yet the pathophysiological mechanisms of host–microbiome interactions and their translation into practice remain insufficiently understood. Because microbiome alterations often reflect latent processes, it is unlikely that a single taxon can serve as a universal biomarker. Instead, polymicrobial shifts are expected, and uncovering these complex patterns represents a central challenge. Latent Dirichlet Allocation (LDA) offers a promising approach but also entails challenges: limited reproducibility, poorly interpretable clusters, restricted causal inference, and lack of generalizability. These issues exacerbate the translational gap in microbiome research - particularly the lack of validation and generalization in independent, population-based, and longitudinal cohorts. To address this gap, this project will develop the C-BIRD framework (Causal Biostatistical Framework for Interpretable and Robust Discovery of Polymicrobial Biomarkers). C-BIRD extends LDA by: (i) averaging across iterations to reduce randomness, (ii) simplifying complex patterns into interpretable key bacteria, (iii) linking results to clinically relevant endpoints, and (iv) applying causal inference to account for systematic biases. The goal is to create a robust, interpretable pipeline that delivers transferable polymicrobial biomarkers for risk stratification, early diagnosis, and prognosis. The application will be demonstrated in systemic lupus erythematosus (SLE) - a multisystemic, inflammatory, and clinically heterogeneous autoimmune disease for which growing evidence suggests a key role of microbiome-immune system interactions. Using diverse clinical longitudinal cohorts, C-BIRD will be applied to identify SLE-relevant polymicrobial markers, link them to serological and clinical phenotypes, and evaluate their reproducibility and generalizability. Furthermore, using the German National Cohort (NAKO), the framework will assess transferability to the general population by testing whether validated microbial SLE biomarkers are detectable in antibody-positive individuals and associated with disease progression and transition to clinically manifest SLE. The project’s goal is methodological advancement, and a deeper understanding of how microbial signatures can be reliably discovered, interpreted, and evaluated across contexts. In doing so, C-BIRD aims to provide a conceptual framework that aligns translational microbiome research more closely with reproducibility and generalizability, thereby paving the way for robust applications in both clinical and population-based research.

• DFG Programme: WBP Fellowship
• International Connection: Australia
• Host: Dr. Simone Li