Bayesian Generalized Linear Models for Analyzing Compositional and Sub‐Compositional Microbiome Data via EM Algorithm DOI
Li Zhang,

Zhenying Ding,

Jinhong Cui

и другие.

Statistics in Medicine, Год журнала: 2025, Номер 44(7)

Опубликована: Март 30, 2025

ABSTRACT The study of compositional microbiome data is critical for exploring the functional roles microbial communities in human health and disease. Recent advances have shifted from traditional log‐ratio transformations covariates to zero constraint on sum corresponding coefficients. Various approaches, including penalized regression Markov Chain Monte Carlo (MCMC) algorithms, been extended enforce this sum‐to‐zero constraint. However, these methods exhibit limitations: yields only point estimates, limiting uncertainty assessment, while MCMC methods, although reliable, are computationally intensive, particularly high‐dimensional settings. To address challenges posed by existing we proposed Bayesian generalized linear models analyzing sub‐compositional data. Our model employs a spike‐and‐slab double‐exponential prior coefficients, inducing weak shrinkage large coefficients strong irrelevant ones, making it ideal handled through soft‐centers applying distribution or subcompositional alleviate computational intensity, developed fast stable algorithm incorporating expectation–maximization (EM) steps into routine iteratively weighted least squares (IWLS) fitting GLMs. performance method was assessed extensive simulation studies. results show that our approach outperforms with higher accuracy coefficient estimates lower prediction error. We also applied one find microorganisms linked inflammatory bowel disease (IBD). implemented freely available R package BhGLM https://github.com/nyiuab/BhGLM .

Язык: Английский

Microbiomes in action: multifaceted benefits and challenges across academic disciplines DOI Creative Commons
Sereyboth Soth, J. G. Hampton, Hossein Alizadeh

и другие.

Frontiers in Microbiology, Год журнала: 2025, Номер 16

Опубликована: Март 18, 2025

Microbiomes combine the species and activities of all microorganisms living together in a specific habitat. They comprise unique ecological niches with influences that scale from local to global ecosystems. Understanding connectivity microbiomes across academic disciplines is important help mitigate climate change, reduce food insecurity, control harmful diseases, ensure environmental sustainability. However, most publications refer individual microbiomes, those integrating two or more related are rare. This review examines multifaceted benefits agriculture, manufacturing preservation, natural environment, human health, biocatalyst processes. Plant by improving plant nutrient cycling increasing abiotic biotic stress resilience, have increased crop yields over 20%. Food generate approximately USD 30 billion economy through fermented industry alone. Environmental detoxify pollutants, absorb than 90% heavy metals, facilitate carbon sequestration. For an adult person can carry up 38 trillion microbes which regulate well being, immune functionality, reproductive function, disease prevention. used optimize processes produce bioenergy biochemicals; bioethanol production alone valued at 83 p.a. challenges, including knowledge gaps, engaging indigenous communities, technical limitations, regulatory considerations, need for interdisciplinary collaboration, ethical issues, must be overcome before potential effectively realized.

Язык: Английский

Процитировано

1
Bayesian Generalized Linear Models for Analyzing Compositional and Sub‐Compositional Microbiome Data via EM Algorithm DOI
Li Zhang,

Zhenying Ding,

Jinhong Cui

и другие.

Statistics in Medicine, Год журнала: 2025, Номер 44(7)

Опубликована: Март 30, 2025

ABSTRACT The study of compositional microbiome data is critical for exploring the functional roles microbial communities in human health and disease. Recent advances have shifted from traditional log‐ratio transformations covariates to zero constraint on sum corresponding coefficients. Various approaches, including penalized regression Markov Chain Monte Carlo (MCMC) algorithms, been extended enforce this sum‐to‐zero constraint. However, these methods exhibit limitations: yields only point estimates, limiting uncertainty assessment, while MCMC methods, although reliable, are computationally intensive, particularly high‐dimensional settings. To address challenges posed by existing we proposed Bayesian generalized linear models analyzing sub‐compositional data. Our model employs a spike‐and‐slab double‐exponential prior coefficients, inducing weak shrinkage large coefficients strong irrelevant ones, making it ideal handled through soft‐centers applying distribution or subcompositional alleviate computational intensity, developed fast stable algorithm incorporating expectation–maximization (EM) steps into routine iteratively weighted least squares (IWLS) fitting GLMs. performance method was assessed extensive simulation studies. results show that our approach outperforms with higher accuracy coefficient estimates lower prediction error. We also applied one find microorganisms linked inflammatory bowel disease (IBD). implemented freely available R package BhGLM https://github.com/nyiuab/BhGLM .

Язык: Английский

Процитировано

0