Adaptive Bayesian Transfer Learning under Compositional Generalized Linear Mixed Model for Microbiome Data Analysis

Background With the development of high-throughput sequencing technologies, microbiome data play 8an increasingly important role in pathological analysis and disease prediction. However, microbiome data are typically characterized by high dimensionality, small sample sizes, and compositional constraints, which present significant challenges to conventional statistical methods, particularly in studies of specific diseases with limited sample sizes. Methods Motivated by these challenges, this paper proposes the Bayesian Transfer Learning under Compositional Generalized Linear Mixed Model and its adaptive extension method. The methods achieve knowledge transfer to the target model by incorporating informative source data and further realize adaptive selection of transfer parameters to prevent negative transfer. It explicitly accounts for compo- sitional constraints inherent in microbiome data, and also incorporates phylogenetic relatedness among taxa and dependence among samples. The performance of the proposed methods is evaluated through comprehensive simulation studies and real data analyses. Results Systematic simulations demonstrate that the proposed methods significantly outperform the baseline method under a variety of source data conditions. In particular, the adaptive transfer strategy enhances the robustness and generalization performance of the method. Applications on real gut microbiome data further validate the effectiveness of the proposed methods. The identified microbial taxa are also consistent with existing biological findings, supporting the interpretability of the proposed methods. Conclusion A Bayesian transfer learning framework is proposed for high-dimensional microbiome data with small sample sizes and compositional constraints. The method mitigates negative transfer through adaptive transfer mechanism, thereby improving robustness and generalization performance, and enhancing phenotype prediction and disease diagnosis.