Modelling Diffuse Subcellular Protein Structures as Dynamic Social Networks

TL;DR

This paper introduces a novel approach to model and analyze the dynamic behavior of diffuse subcellular structures like mitochondria using probabilistic mixtures and graph theory, providing a rigorous framework for bioimaging analysis.

Contribution

It presents a new method to model subcellular morphologies as dynamic social networks using probabilistic mixtures and graph Laplacians, enabling quantitative analysis.

Findings

Effective modeling of mitochondrial patterns as dynamic networks

Graph analysis yields biologically meaningful insights

Framework applicable to other diffuse subcellular structures

Abstract

Fluorescence microscopy has led to impressive quantitative models and new insights gained from richer sets of biomedical imagery. However, there is a dearth of rigorous and established bioimaging strategies for modeling spatiotemporal behavior of diffuse, subcellular components such as mitochondria or actin. In many cases, these structures are assessed by hand or with other semi-quantitative measures. We propose to build descriptive and dynamic models of diffuse subcellular morphologies, using the mitochondrial protein patterns of cervical epithelial (HeLa) cells. We develop a parametric representation of the patterns as a mixture of probability masses. This mixture is iteratively perturbed over time to fit the evolving spatiotemporal behavior of the subcellular structures. We convert the resulting trajectory into a series of graph Laplacians to formally define a dynamic network. Finally, we demonstrate how graph theoretic analyses of the trajectories yield biologically-meaningful quantifications of the structures.