CsppaRshiny: An interactive R/Shiny App for exploration of cellular spatial point patterns

Overview

The R shiny app CsppaRshiny performs machine learning analysis and visualization of cellular spatial point patterns under hypercaloric diet shifts and it is based on Csppa R-package.

The application performs 3D and 2D for visualization of spatial point pattern density plots, allowing flexibility in order to represent the data and emphasize the question of interest. Further, it allows for performing overall and local significance analysis of spatial point pattern densities, employing several statistical approaches.

K Nearest Neighbour and Random Forest classification algorithms are implemented to compare the grouping of the cells expressing different markers within and between the diets. On top of this, correlation and spatial auto-correlation of the cells expressing different markets can be compared using Mentel and Moran I test, respectively.

Application

Here we focus on the astrocytes from the arcuate nucleus of the mouse brain and the expression of Gfap and Aldh1l1 genes recovering spatial point patterns under standard chow (SC), 5 and 15 days high-fat high sugar (HFHS) diet. The R-package Csppa allows for assessing whether these astrocyte populations are spatially organized and whether they tend to form local identical clusters in response to an HFHS diet over time. To do that, the algorithm measures the degree of spatial coherence (depicting the level of similarity between neighbors) of each astrocytic sub-type in different conditions (SC, 5d, or 15d HFHS diet) by applying Moran I spatial autocorrelation coefficient, previously described as an indicator of the level of spatial dispersion. On top of that, employing a random forest classifier determines the partitioning of the feature space shared by astrocytes expressing Gfap and Aldh1l1 in each experimental group.