Graph kernel methods for graph structure/topology comparisons

Performs graph structural comparisons using graph kernel-based method. Currently available kernel methods include: the Weisfeiler-Lehman kernel, the graphlet kernel, and the random walk kernel.

AntibodyForests_kernels(
  trees,
  graph.type,
  kernel.method,
  additional.param,
  max.networks
)

Arguments

trees

(nested) list of AntibodyForests objects, as obtained from the AntibodyForests function, to be compared.

graph.type

string - 'tree' will use the usual output of the AntibodyForests function (tree graphs), 'heterogeneous' will use the output of the AntibodyForests_heterogeneous function (bipartite networks for both cells and sequences).

kernel.method

string - kernel method to be used, as implemented in the 'graphkernels' R package. 'weisfeiler_lehman' for the Weisfeiler-Lehman kernel, 'graphlet, and 'random_walk'.

additional.param

integer - additional kernel options/parameters (e.g., kernel iterations for the Weisfeiler-Lehman kernel).

max.networks

integer - maximum number of networks to be compared (will pick the networks with the most number of cells first).

Value

Heatmap of the graph kernel values.

Examples

if (FALSE) {
AntibodyForests_kernels(trees, graph.type = 'tree',
kernel.method = 'weisfeiler_lehman',
additional.params = 10, max.networks = 50)
}