Using SpatialDE to characterize spatiotemporal changes in mitochondrial morphology

Brittany Dorsey, a sophomore from Mercer University, worked with Dr. Shannon Quinn and Dr. Fred Quinn to test the use of a new method to detect changes in organelle morphology.

Abstract:   Intracellular bacterial pathogens have the capacity to greatly alter target organelles’ morphology, which can easily be visualized through fluorescence microscopy, but is more difficult to quantify succinctly. Work is being done to consider Gaussian Mixture Models as a viable solution by viewing mitochondria as social networks, but there are difficulties with this method. Therefore, the goal of the current project is to explore the feasibility of SpatialDE as an alternative way to quantify the spatiotemporal changes in organelle morphology. Using time series footage of mitochondria, three morphological phenotypes were analyzed: control, fragmented, and fused. The raw video was converted to a three-dimensional matrix of pixel values, which was then raster scanned into a two-dimensional matrix. This matrix was normalized, then input into the SpatialDE framework using the programming language Python. The data frame output gave 18 different variable values for each pixel location throughout the footage, which was converted back into “image” format in order to be analyzed. The results showed little to no discernable patterns between treatments. In comparison, the Gaussian Mixture Model output shows clear similarities among phenotypes. Therefore, it was determined that Gaussian Mixture Models continue to be the best option to model spatiotemporal changes in diffuse organelles. Once this method is fully developed, the mechanisms by which bacterial virulence factors transform mitochondrial structure in host cells will be better understood, which will have crucial implications for structural biology and biomedicine.


Download (PDF, 1.85MB)