This function optimizes the choice of ClusterONE algorithm
parameters such as density, node penalty, and overlap score by comparing
clustering-derived partitions for each combination of parameters to known
labels (i.e., CORUM complexes) and assess the similarity between them
using quality measures including overlap score, sensitivity (Sn),
clustering-wise positive predictive value (PPV), geometric accuracy (Acc),
and maximum matching raio (MMR).It is recommended to first reduce
redundancy in the known reference complexes via
EliminateCpxRedundance,then performs parameter tuning.
cluster_tuning(
refcpx,
csize = 2,
d = c(0.3, 0.5),
p = c(2),
max_overlap = c(0.5, 0.6),
tpath = file.path(system.file("extdata", package = "MACP"))
)A data.frame containing clustering performance across different combination of parameters.
A list containing reference complexes (i.e., corum complexes).
An integer, the minimum size of the predicted complexes. Defualts to 2.
A vector of number, density of predicted complexes.
A vector of integer, penalty value for the inclusion of each node.
A vector of number, specifies the maximum allowed overlap between two clusters.
A character string indicating the path to the project directory that contains the interaction data. Interaction data must be stored as .txt file and containing id1-id2-weight triplets. Defaults to MACP/inst/extdata directory.
Matineh Rahmatbakhsh, matinerb.94@gmail.com
cluster_tuning
Nepusz, T., Yu, H., and Paccanaro, A. (2012a). Detecting overlapping protein complexes in protein-protein interaction networks. Nat. Methods 9, 471.