Skip to contents

Generates CoReMo-based gene modules

cor_module_coremo_clustering.Rd

This function creates gene modules, based on the framework from Srivastava et al., 2018. Briefly, it applies an RBF function to the correlation matrix to reduce the impact of weak correlations, leverages hierarchical clustering for clustering the genes. It optimises the R2 (cor^2) within the clusters to identify the optimal cut. Gene modules with low R2 are being considered as the 'junk module'.

Usage

cor_module_coremo_clustering(
  object,
  span = 0.1,
  coremo_params = params_coremo(),
  seed = 42L,
  .verbose = TRUE
)

Arguments

object

The class, see BulkCoExp().

span

Float. Defines the span parameter for the loess function to identify the inflection point. Defaults to 0.1. Must be between 0.1 and 1.

coremo_params

List. Parameters for the generation of the CoReMo modules, see params_coremo(). Contains:

  • epsilon - Float. The epsilon parameter for the RBF. You can optimise that one with cor_module_check_epsilon(). Defaults to 2.

  • k_min - Integer. Minimum number of cuts. Defaults to 2L.

  • k_max - Integer. Maximum number of cuts. Defaults to 150L.

  • min_size - Optional integer. Minimum size of the clusters. If provided, smaller clusters will be merged by eigengene similarity.

  • junk_module_threshold - Float. Minimum R2 median value for a module to not be considered a junk module. Defaults to 0.05.

  • rbf_func - String. Type of RBF function to apply. Defaults to "gaussian".

  • cor_method - String. Type of correlation method to use for merging the smaller cluster. Defaults to "spearman".

seed

Integer. Random seed for reproducibility purposes.

.verbose

Boolean. Controls verbosity of the function.

Value

The class with added data to the properties for subsequent usage.

References

Srivastava, et al., Nat. Commun., 2018