A Bayesian Multivariate Spatial Point Pattern Model: Application to Oral Microbiome FISH Image Data.

Area of Science:

• Microbiology
• Computational Biology
• Biostatistics

Background:

• Cellular imaging, particularly fluorescence in situ hybridization (FISH), enables detailed visualization of cell spatial organization.
• Quantifying this organization is vital for understanding tissue and biofilm function, impacting human health and disease.
• Existing methods lack comprehensive approaches for quantifying complex multi-cell spatial interactions.

Purpose of the Study:

• To develop a flexible multivariate point process model for characterizing and estimating spatial interactions among multiple cell types.
• To provide a Bayesian framework for unified estimation and direct quantification of uncertainty in spatial relationships.
• To enable robust model selection and hierarchical inference for integrated analysis of image data.

Main Methods:

• Proposed a flexible multivariate point process model within a Bayesian framework.
• Incorporated shrinkage priors for stable and interpretable estimation of latent processes.
• Utilized a deviance information criterion for model selection and comparison with latent variables.
• Developed a hierarchical modeling approach to integrate multiple image-specific estimates.

Main Results:

• Identified strong positive spatial correlations between specific bacterial pairs in human tongue biofilms (e.g., Streptococcus mitis-Streptococcus salivarius).
• Observed negative spatial correlations for other pairs (e.g., Actinomyces-Rothia).
• Demonstrated that inter-taxon relationships significantly contribute to the spatial variance for most taxa.

Conclusions:

• The proposed multivariate point process model effectively quantifies complex spatial interactions in microbial communities.
• The findings reveal specific spatial relationships among oral bacteria, offering insights into biofilm structure and function.
• This quantitative approach has implications for understanding host-microbe interactions and developing targeted interventions.