Point process models for localization and interdependence of punctate cellular structures.

Area of Science:

• Cellular and Molecular Biology
• Computational Biology
• Biophysics

Background:

• Accurate cellular organization models are crucial for predicting dynamic cellular functions.
• The CellOrganizer platform generates cellular component models from microscopy images, capturing heterogeneity.
• Previous models focused on punctate organelle patterns relative to cell and nuclear membranes, and microtubules.

Purpose of the Study:

• To extend organelle localization models to include the endoplasmic reticulum (ER) and inter/intra-punctate relationships.
• To develop classifiers for updating protein assignments to specific localization patterns.
• To assess the utility of spatial point process models for understanding cellular structure interactions.

Main Methods:

• Developed non-parametric models for punctate organelle patterns.
• Extended models to incorporate relationships with the endoplasmic reticulum (ER) and other puncta of the same type.
• Built classifiers to assign proteins to 11 distinct localization patterns across three cell lines.

Main Results:

• Punctate patterns examined were not found to be dependent on ER position or inter/intra-class proximity.
• Generative models successfully created statistically accurate representations of puncta localization using cellular markers.
• The approach captured complex cellular structure interactions with minimal human input.

Conclusions:

• Spatial point process models offer valuable insights into spatial dependence between cellular structures.
• The refined CellOrganizer models provide a novel approach to vesicular protein annotation in high-throughput microscopy.
• Accurate generative models of cellular organization are essential for predictive modeling of cellular dynamics.