Cell Group Construction • CellTFusion

library(CellTFusion)
#> 
#>

Cell group construction is the core step of the CellTFusion pipeline. Using previously computed features (TF modules, deconvolution groups, and pathway activities), we identify clusters of samples that share similar biological activity patterns as reflected by TF module scores.

This step requires the outputs from the Feature Computation article:

counts.norm — log-normalized expression matrix
tfs — TF activity matrix
deconv — cell-type deconvolution proportions
network — WTCNA TF modules
dt — reduced deconvolution groups
traitdata — clinical metadata

Supervised cell group construction

In the supervised mode, clinical traits guide the identification of cell groups associated with specific phenotypes (e.g., treatment response). A trait column and its positive class must be provided:

cell_groups <- construct_cell_groups(
  counts.norm, tfs, deconv, network, dt,
  traitdata,
  pval     = 0.05,
  trait    = "Best.Confirmed.Overall.Response",
  positive = "CR"
)

Unsupervised cell group construction

Alternatively, cell groups can be identified purely from molecular features, without using clinical annotations:

cell_groups <- construct_cell_groups(
  counts.norm, tfs, deconv, network, dt,
  pval = 0.05
)

Output structure

construct_cell_groups() returns a list with three elements:

Element	Description
`[[1]]`	Data frame of cell group scores (samples × groups)
`[[2]]`	Named list of cell-type compositions per group
`[[3]]`	Named list of loading vectors per group

Results are also written to Results/Cell.groups.composition.csv and Results/Cell.groups.scores.csv.