In jafarilab/NIMAA: Nominal Data Mining Analysis

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

NIMAA

The NIMAA package [@nimaa] provides a comprehensive set of methods for performing nominal data mining.

It employs bipartite networks to demonstrate how two nominal variables are linked, and then places them in the incidence matrix to proceed with network analysis. NIMAA aids in characterizing the pattern of missing values in a dataset, locating large submatrices with non-missing values, and predicting edges within nominal variable labels. Then, given a submatrix, two unipartite networks are constructed using various network projection methods. NIMAA provides a variety of choices for clustering projected networks and selecting the best one. The best clustering results can also be used as a benchmark for imputation analysis in weighted bipartite networks.

Installation

You can install the released version of NIMAA from CRAN with:

install.packages("NIMAA")

And the development version from GitHub with:

# install.packages("devtools")
devtools::install_github("jafarilab/NIMAA")

Example

Plotting the original data

library(NIMAA)
## load the beatAML data
beatAML_data <- NIMAA::beatAML

# plot the original data
beatAML_incidence_matrix <- plotIncMatrix(
  x = beatAML_data, # original data with 3 columns
  index_nominal = c(2,1), # the first two columns are nominal data
  index_numeric = 3,  # the third column is numeric data
  print_skim = FALSE, # if you want to check the skim output, set this as TRUE
  plot_weight = TRUE, # when plotting the weighted incidence matrix
  verbose = FALSE # NOT save the figures to local folder
  )

knitr::include_graphics("vignettes/patient_id-inhibitor.png")

Plotting the bipartite network of the original data

plotBipartite(inc_mat = beatAML_incidence_matrix, vertex.label.display = T)

Extracting large submatrices without missing values

The extractSubMatrix() function extracts the submatrices that have non-missing values or have a certain percentage of missing values inside (not for elements-max matrix), depending on the argument's input. The package vignette and help manual contain more details.

sub_matrices <- extractSubMatrix(
  x = beatAML_incidence_matrix,
  shape = c("Square", "Rectangular_element_max"), # the selected shapes of submatrices
  row.vars = "patient_id",
  col.vars = "inhibitor",
  plot_weight = TRUE,
  print_skim = FALSE
  )

knitr::include_graphics("vignettes/Row_wise_arrangement.png")

knitr::include_graphics("vignettes/Column_wise_arrangement.png")

Cluster finding analysis of projected unipartite networks

The findCluster() function implements seven widely used network clustering algorithms, with the option of preprocessing the input incidence matrix following the projecting of the bipartite network into unipartite networks. Also, internal and external measurements can be used to compare clustering algorithms. Details can be found in the package vignette and help manual.

cls <- findCluster(
  sub_matrices$Rectangular_element_max,
  part = 1,
  method = "all", # all available clustering methods
  normalization = TRUE, # normalize the input matrix
  rm_weak_edges = TRUE, # remove the weak edges in the network
  rm_method = 'delete', # delete the weak edges instead of lowering their weights to 0.
  threshold = 'median', # Use median of edges' weights as threshold
  set_remaining_to_1 = TRUE, # set the weights of remaining edges to 1
  )

Edge predicting in weighted bipartite networks

The predictEdge() function predicts new edges between nominal variables' labels or imputes missing values in the input data matrix using several imputation methods. We can compare the imputation results using the validateEdgePrediction() function to choose the best method based on a predefined benchmark. The package vignette and help manual contain more details.

imputations <- predictEdge(
  inc_mat = beatAML_incidence_matrix,
  method = c('svd','median','als','CA')
  )

validateEdgePrediction(imputation = imputations,
                   refer_community = cls$fast_greedy,
                   clustering_args = cls$clustering_args)