Derive communication features from a kernel

Derive communication features from a kernel

Usage

compute_kernel_features(
  kernel,
  unifKernel = NULL,
  celltypes,
  communication_type = "D",
  bundle = TRUE,
  patient_names = NULL,
  Dcell = NULL,
  norm = FALSE,
  patient_idx = NULL,
  zero_threshold = 1
)

Arguments

kernel

Kernel array from compute_kernel().

unifKernel

Optional normalized baseline kernel.

celltypes

Character vector of cell-type labels.

communication_type

Feature family to compute ("D", "W", "TT", or "GSCC"; "TT" returns both trust- and cycle-triangle columns together, see computeTriangles()), or a character vector of several of these. Since the kernel itself (kernel/unifKernel) is passed in already computed, requesting multiple types here costs nothing extra to compute per type beyond that one kernel – no re-derivation happens for any of them. With more than one type, the return value is a named list (one data frame per type) instead of a single data frame.

bundle

Logical; if TRUE, merge directionally equivalent features where appropriate.

patient_names

Optional patient labels.

Dcell

Patient-by-cell-type abundance matrix. Always required for "GSCC"; required for "D", "W", "TT" only when unifKernel is not supplied (raw, unnormalized features).

norm

Logical; if TRUE, compute normalized features when a baseline is supplied.

patient_idx

Optional patient index subset.

zero_threshold

Drop a feature once its fraction of zero-valued patients reaches this threshold. Default 1 only drops features that are zero for every patient (the original behavior); lower it (e.g. 0.9) to also drop merely zero-inflated features. Passed through to calculateDirect()/calculateWedges()/computeTriangles(); ignored for "GSCC".

Value

A data frame of feature values for the selected patients, or (when communication_type has more than one entry) a named list of such data frames, one per requested type.