Abstract
Gut microbiome plays a significant role in defining the health status of subjects, and recent studies highlight the importance of using time series strategies to analyse microbiome dynamics. In this paper, we develop a Bayesian model for microbiota longitudinal data, based on Dirichlet distribution with time‐varying parameters, that take into account the compositional paradigm and consider principal balances. The proposed model can be effective for predicting the future dynamics of a microbial community in the short term and for analysing the microbial interactions using the value of the estimated parameters. The usefulness of the proposed model is illustrated with six different datasets, and a comparison study with four alternative models is provided.
Funding Information
  • Ministerio de Economía y Competitividad (MTM2017-83850-P, SAF2012-31187, SAF2013-49788-EXP, SAF2015-65878-R)
  • Instituto de Salud Carlos III (PIE14/00045, AC15/00022)
  • Generalitat Valenciana (PrometeoII/2014/065, Prometeo/2018/A/133)
  • European Regional Development Fund

This publication has 54 references indexed in Scilit: