Network Visualizations

[GWMcElfresh]

TODO on fleshing this out, but the primary goals are to extend this package into network visualizations and analysis.

Example visualization:

This network establishes a graph where we can care primarily about two quantities:

• what is the probability that node of type X (controller) is connected to a node of type Y (progressor).
• is the typical distance between controllers vs progressors different?

To that end, this now supports on-the-fly network construction via something like this (for the above network):

tcrClustR::TCRDistanceNetwork(
  seuratObj = seuratObjTcrA, 
  chains = "TRA",
  colorBy = "outcome", 
  edgeType = 'continuous'
)

and then a modeling suite for connectivity vs distances like so:

result_diana <- tcrClustR::ModelTCRDyads(
  seuratObj_TCR        = seuratObjTcrA,
  chains               = "TRA",
  groupColumn          = "outcome",
  estimand = 'effects', 
  distanceThreshold    = 50,
  communityMethod      = "DIANA",
  clusterSizeThreshold = 2,
  subjectIdCol         = "SubjectId",
  verbose              = TRUE
)

This modeling result will compute the connectivity and the distance results and store them, so they can be accessed by visualization functions. The models are either estimand = effects or estimand = means, which equates to either a contrast across values or the values of groupColumn themselves.

For instance, for estimand = effects we contrast and plot like so:

tcrClustR::PlotDyadEstimates(result_diana, which = "connectivity")

tcrClustR::PlotDyadEstimates(result_diana, which = "distance")

but if we run with estimand = 'means':

tcrClustR::PlotDyadEstimates(result_diana, which = "connectivity")

tcrClustR::PlotDyadEstimates(result_diana, which = "distance")

Specifically for the conditional connectivity probabilities, this function subsets unpaired nodes (e.g. clones with high distance from all other clones). So, by passing showUnpaired = TRUE, we recover those probabilities in a facet.

tcrClustR::PlotDyadEstimates(result_diana, which = "connectivity", showUnpaired = TRUE)