Multivariate Cluster Point Process to Quantify and Explore Multi-Entity Configurations: Application to Biofilm Image...

Area of Science:

• Spatial statistics
• Computational biology
• Biostatistics

Background:

• Analyzing spatial relationships of objects, such as cells in biomedical images, is crucial for understanding biological mechanisms.
• Traditional methods often overlook the central object's role in cluster formation, limiting analysis of complex arrangements.
• Parent-offspring frameworks offer potential but require adaptation for multi-object systems.

Purpose of the Study:

• To introduce a novel multivariate cluster point process (MCPP) for quantifying multi-object spatial arrangements.
• To develop a statistical framework that accounts for multilayered and multivariate clustering, utilizing central 'parent' object locations.
• To compare the MCPP's performance against existing models using simulated and real-world biological data.

Main Methods:

• Development of the multivariate cluster point process (MCPP) model, specifying parent and offspring object types.
• Utilizing the deviance information criterion (DIC) to compare model fits and explore unknown roles of object types.
• Validation using simulated data to assess accuracy and precision against the Neyman-Scott process model.
• Application to human dental plaque biofilm image data to quantify known and discover novel spatial relationships.

Main Results:

• The MCPP accurately identified simulated spatial relationships and provided more precise parameter estimates than the univariate Neyman-Scott process model.
• In dental plaque data, MCPP quantified clustering of Streptococcus and Porphyromonas around Corynebacterium, and Pasteurellaceae around Streptococcus.
• The model successfully captured hypothesized structures and suggested a novel clustering relationship between Fusobacterium and Leptotrichia.

Conclusions:

• The multivariate cluster point process (MCPP) is an effective tool for quantifying complex multi-object spatial arrangements in biological systems.
• MCPP facilitates the discovery of novel biological interactions by accurately modeling spatial configurations.
• This approach enhances our understanding of micro-environmental structures and their implications in fields like microbiology and cell biology.