Background
A typical study design will contrast several disease or treated samples with one or more controls. In these studies, it can be difficult to identify the (interesting) ways in which the disease samples are perturbed compared to the controls. To this end, we have created an R package, “InSituDiff”, containing a suite of tools for exploring how cellular neighborhoods are perturbed in disease compared to controls.
The heart of InSituDiff is the measurement of how “cellular neighborhoods” (a target cell and its nearest neighbors) contrast with their most similar control cellular neighborhood. This analysis produces a perturbation score for all cells * genes. With these perturbation scores in hand, numerous useful analyses become possible:
- Spatial maps of gene perturbation scores are often more informative than maps of expression level.
- We can identify highly perturbed genes, and we can quantify the spatial dependence of their perturbations.
- By applying spatial clustering algorithms to the perturbation matrix, we can discover distinct impacts of disease across the span of tissues.
- By clustering genes’ perturbation values, we discover sets of genes with spatially correlated perturbations.
The above suite of analyses very quickly brings the major impacts of disease into focus. Conveniently, this analysis is agnostic to your cell typing results, making it a plausible tool for first-pass data exploration.
Resources
See the package vignette for a demonstration of the workflow.