@@ -32,7 +32,8 @@ Depends:

     clusterCrit (>= 1.2.8),

     clValid (>= 0.7),

     glmnet (>= 4.1.3),

-    dplyr(>= 1.0.7)

+    dplyr(>= 1.0.7),

+    stats (>= 4.1.2)

 Suggests:

     rmarkdown,

     knitr,

@@ -76,3 +76,15 @@ export(plot_feature_selection)

 export(plot_partition_agreement)

 export(plot_signature_feature)

 export(plot_vote_frequencies)

+import(ggplot2)

+importFrom(clValid,clusters)

+importFrom(diceR,prepare_data)

+importFrom(dplyr,"%>%")

+importFrom(dplyr,across)

+importFrom(dplyr,filter)

+importFrom(dplyr,left_join)

+importFrom(fpc,hclustCBI)

+importFrom(fpc,kmeansCBI)

+importFrom(fpc,speccCBI)

+importFrom(pdfCluster,adj.rand.index)

+importFrom(stats,cutree)

new file mode 100644

@@ -7,9 +7,10 @@

 #' @param algorithm The clustering algorithm to use for the multiple clustering

 #' runs to be measured

 #'

-#' @return An object of class "clusterVoting" containing a matrix with metric scores for every k and

-#' internal index, cluster memberships for every k, a dataframe with the k votes

-#'  for every index, k vote frequencies and the frequency barplot of the k votes

+#' @return An object of class "clusterVoting" containing a matrix with metric

+#'  scores for every k and internal index, cluster memberships for every k, a

+#'  dataframe with the k votes for every index, k vote frequencies and the

+#'  frequency barplot of the k votes

 #'

 #' @export

 #'

@@ -17,6 +18,9 @@

 #' clusterVoting(toy_genes, 4,14,"sc")

 #' clusterVoting(toy_genes, 2,7,"hc")

 #' clusterVoting(toy_genes, 2,4,"km")

+#'

+#' @importFrom diceR prepare_data

+#' @import ggplot2

 clusterVoting <- function(data ,min.k ,max.k, algorithm) {

@@ -45,24 +49,25 @@ clusterVoting <- function(data ,min.k ,max.k, algorithm) {

   for(current_k in min.k:max.k) {

     if(algorithm == "sc") {

-      cl <- specc(data, centers=current_k, kernel = "rbfdot")

+      cl <- kernlab::specc(data, centers=current_k, kernel = "rbfdot")

       cls <- [email protected]

     } else if(algorithm == "hc") {

-      dist_mat <- dist(data, method = "euclidean")

-      cl <- hclust(dist_mat, method = "average")

-      cls <- cutree(cl, k = current_k)

+      dist_mat <- stats::dist(data, method = "euclidean")

+      cl <- stats::hclust(dist_mat, method = "average")

+      cls <- stats::cutree(cl, k = current_k)

     } else if(algorithm == "km") {

-      cl <- kmeans(data, current_k, algorithm = "Hartigan-Wong")

+      cl <- stats::kmeans(data, current_k, algorithm = "Hartigan-Wong")

       cls <- cl$cluster

     }

-    criteria <- intCriteria(data,cls,c("calinski_harabasz","dunn","pbm","tau",

-                                       "gamma", "c_index","davies_bouldin",

+    criteria <- clusterCrit::intCriteria(data,cls,c("calinski_harabasz","dunn",

+                                                    "pbm","tau", "gamma",

+                                                    "c_index","davies_bouldin",

                                        "mcclain_rao","sd_dis", "ray_turi",

                                        "g_plus","silhouette","s_dbw"))

     con <- clValid::connectivity(clusters = cls, Data = data)

-    comp <- compactness(data, cls)

+    comp <- diceR::compactness(data, cls)

     criteria <- c(criteria, connectivity=con, compactness=comp)

     criteria <- array(as.numeric(unlist(criteria)))

     scores[, counter] <- criteria

@@ -104,7 +109,8 @@ clusterVoting <- function(data ,min.k ,max.k, algorithm) {

   colnames(ensemble.results) <- c("k", "Frequency")

   ensemble.results$Frequency <- as.numeric(ensemble.results$Frequency)

-  ensemble.plot <- ggplot(ensemble.results, aes(k, Frequency, fill = k)) +

+  ensemble.plot <- ggplot2::ggplot(ensemble.results, aes(k, Frequency,

+                                                         fill = k)) +

     geom_col() +

     scale_fill_brewer(palette="Dark2")

@@ -16,6 +16,9 @@

 #' number.of.comparisons = 4)

 #' clusteringMethodSelection(toy_genes, method.upper.k = 2,

 #' number.of.comparisons = 2)

+#'

+#' @import ggplot2

+#' @importFrom clValid clusters

 clusteringMethodSelection <- function(data, method.upper.k = 5,

                                       number.of.comparisons = 3) {

@@ -114,10 +117,10 @@ clusteringMethodSelection <- function(data, method.upper.k = 5,

   s.mean <- mean(df.final$spectral)

   k.mean <- mean(df.final$kmeans)

-  df.plot <- melt(df.final, id=c("clusters"))

+  df.plot <- reshape::melt(df.final, id=c("clusters"))

   colnames(df.plot) <- c("clusters", "methods", "value")

-  agreements.plot <- ggplot(df.plot, aes(x = clusters, y = value)) +

+  agreements.plot <- ggplot2::ggplot(df.plot, aes(x = clusters, y = value)) +

     geom_line(aes(color = methods)) +

     geom_hline(aes(yintercept=h.mean), linetype="dashed") +

     geom_hline(aes(yintercept=s.mean), linetype="dashed") +

@@ -15,6 +15,8 @@

 #' @examples

 #' feasibilityAnalysisDataBased(data = toy_genes, classes = 3)

 #' feasibilityAnalysisDataBased(data = toy_genes, classes = 2)

+#'

+#' @importFrom fpc speccCBI

 feasibilityAnalysisDataBased <- function(data, classes = 3) {

   samples = dim(data)[1]

@@ -39,7 +41,7 @@ feasibilityAnalysisDataBased <- function(data, classes = 3) {

   cl.index <- 1

   feature.index <- 1

   for (i in 1:features) {

-    temp <- rnorm(n = samples,

+    temp <- stats::rnorm(n = samples,

                   mean = c(class.means),

                   sd = class.sd)

     dataset[, ncol(dataset) + 1] <- temp # Append temp column

@@ -68,7 +70,7 @@ feasibilityAnalysisDataBased <- function(data, classes = 3) {

   }

   for (rep in c.min:c.max) {

-    sc.boot <- clusterboot(

+    sc.boot <- fpc::clusterboot(

       stability.dataset,

       B = 25,

       bootmethod = "boot",

@@ -16,6 +16,8 @@

 #' @examples

 #' feasibilityAnalysis(classes = 3, samples = 320, features = 400)

 #' feasibilityAnalysis(classes = 4, samples = 400, features = 120)

+#'

+#' @importFrom fpc speccCBI

 feasibilityAnalysis <- function(classes = 3, samples = 320, features = 400) {

@@ -29,7 +31,7 @@ feasibilityAnalysis <- function(classes = 3, samples = 320, features = 400) {

   cl.index <- 1

   feature.index <- 1

   for(i in 1:features) {

-    temp <- rnorm(n = samples, mean = c(class.means), sd = class.sd)

+    temp <- stats::rnorm(n = samples, mean = c(class.means), sd = class.sd)

     dataset[ , ncol(dataset) + 1] <- temp # Append temp column

     colnames(dataset)[ncol(dataset)] <- paste0("feature_", feature.index)

     feature.index <- feature.index + 1

@@ -56,7 +58,7 @@ feasibilityAnalysis <- function(classes = 3, samples = 320, features = 400) {

   for(rep in c.min:c.max) {

-    sc.boot <- clusterboot(stability.dataset,

+    sc.boot <- fpc::clusterboot(stability.dataset,

                            B = 25,

                            bootmethod = "boot",

                            clustermethod = speccCBI,

@@ -16,6 +16,10 @@

 #' featureSelection(toy_genes, min.k = 3, max.k = 9, step = 3)

 #' featureSelection(toy_genes, min.k = 2, max.k = 4, step = 4)

 #'

+#' @importFrom fpc speccCBI

+#' @import ggplot2

+

+

 featureSelection <- function(data, min.k = 2, max.k = 4, step = 5) {

   print("Selecting feature subset...")

@@ -27,7 +31,7 @@ featureSelection <- function(data, min.k = 2, max.k = 4, step = 5) {

   averages.of.all.k <- list()

   # Sorted features based on variance across data points

-  features.variance <- data.frame(apply(data, 2, var))

+  features.variance <- data.frame(apply(data, 2, stats::var))

   colnames(features.variance) <- "variance"

   sorted.features.variance <-

     features.variance[order(features.variance$variance,decreasing = TRUE), ,

@@ -47,7 +51,7 @@ featureSelection <- function(data, min.k = 2, max.k = 4, step = 5) {

       cur <- sorted.features.variance$names[1:fs]

-      sc.boot <- clusterboot(data[,cur],

+      sc.boot <- fpc::clusterboot(data[,cur],

                              B = 25,

                              bootmethod = "boot",

                              clustermethod = speccCBI,

@@ -85,7 +89,7 @@ featureSelection <- function(data, min.k = 2, max.k = 4, step = 5) {

   all.feature.k.stabilities$featureSet <-

     as.integer(as.character(all.feature.k.stabilities$featureSet))

-  stabilities.plot <- ggplot(data = all.feature.k.stabilities,

+  stabilities.plot <- ggplot2::ggplot(data = all.feature.k.stabilities,

                              aes(x=featureSet, y=means)) +

     geom_line(color='firebrick',group = 1, size = 0.5) +

     geom_point(color='firebrick', group = 1) +

@@ -14,9 +14,15 @@

 #'

 #' @examples

 #' geneSignatures(toy_genes, toy_gene_memberships)

+#'

+#' @import ggplot2

+#' @importFrom dplyr across filter %>% left_join

+

 geneSignatures <- function(data, memberships) {

+    # utils::globalVariables("where", add=FALSE)

+

     data <- as.data.frame(data)

     rnames <- row.names(data)

     data$id <- rnames

@@ -39,19 +45,19 @@ geneSignatures <- function(data, memberships) {

     # Running cross-validation Lasso to find optimal lambda value

     data.matrix <- as.matrix(data[,2:dim(data)[2]])

-    cv_model <- cv.glmnet(data.matrix(data.matrix), data$membership, family = "multinomial",

-                          alpha = 1)

+    cv_model <- glmnet::cv.glmnet(data.matrix(data.matrix), data$membership,

+                                  family = "multinomial", alpha = 1)

     # Optimal lambda value (minimizing test MSE)

     optimal_lambda <- cv_model$lambda.min

     # Running optimal lasso model

-    optimal_lasso <- glmnet(data.matrix(data.matrix), data$membership,

+    optimal_lasso <- glmnet::glmnet(data.matrix(data.matrix), data$membership,

                             family = "multinomial",

                             alpha = 1, lambda = optimal_lambda)

     # Extract coefficients for minimized test MSE)

-    Coefficients <- coef(optimal_lasso, s = "min")

+    Coefficients <- stats::coef(optimal_lasso, s = "min")

     # Formatting coefficient dataframe per cluster

     ns <- names(Coefficients)

@@ -78,9 +84,10 @@ geneSignatures <- function(data, memberships) {

     # retain top 30%

     coef.dataset <- coef.dataset[1:round(dim(coef.dataset)[1]*0.3, digits = 0),]

-    coef.data.melt <- melt(coef.dataset)

+    coef.data.melt <- reshape::melt(coef.dataset)

-    coef.30perc <- ggplot(data = coef.data.melt, aes(x = features, y = value,

+    coef.30perc <- ggplot2::ggplot(data = coef.data.melt,

+                                   aes(x = features, y = value,

                                                      fill = variable)) +

         geom_bar(stat = "identity") +

         theme(axis.title.x=element_blank(),

@@ -13,6 +13,8 @@

 #' @examples

 #' optimalClustering(toy_genes, 4,"spectral")

 #' optimalClustering(toy_genes, 2,"kmeans")

+#'

+#' @importFrom fpc speccCBI hclustCBI kmeansCBI

 optimalClustering <- function(data, clusters, algorithm) {

@@ -28,7 +30,7 @@ optimalClustering <- function(data, clusters, algorithm) {

     for (par in spectral.kernels) {

-      sc.boot <- clusterboot(data,

+      sc.boot <- fpc::clusterboot(data,

                              B = 25,

                              bootmethod = "boot",

                              clustermethod = speccCBI,

@@ -55,7 +57,7 @@ optimalClustering <- function(data, clusters, algorithm) {

     for (par in hierarchical.methods) {

-      sc.boot <- clusterboot(data,

+      sc.boot <- fpc::clusterboot(data,

                              B = 25,

                              bootmethod = "boot",

                              clustermethod = hclustCBI,

@@ -81,7 +83,7 @@ optimalClustering <- function(data, clusters, algorithm) {

     for (par in kmeans.kernels) {

-      sc.boot <- clusterboot(data,

+      sc.boot <- fpc::clusterboot(data,

                              B = 25,

                              bootmethod = "boot",

                              clustermethod = kmeansCBI,

@@ -45,6 +45,9 @@

 #' partitionAgreement(toy_genes, algorithm.1 = "spectral", measure.1 = "rbfdot",

 #' algorithm.2 = "kmeans",measure.2 = "Lloyd", number.of.clusters = 5)

 #' @export

+#'

+#' @importFrom stats cutree

+#' @importFrom pdfCluster adj.rand.index

 partitionAgreement <- function(data, algorithm.1 = "hierarchical",

                                measure.1 = "canberra",

@@ -62,22 +65,22 @@ partitionAgreement <- function(data, algorithm.1 = "hierarchical",

   for(i in 2:number.of.clusters) {

     #Spectral clustering

     if(algorithm.1 == "spectral") {

-      cl <- specc(dataset, centers=i, kernel = measure.1)

+      cl <- kernlab::specc(dataset, centers=i, kernel = measure.1)

       temp <- data.frame([email protected])

       cr1 <- cbind(cr1, temp)

     }

     #Hierarchical

     else if(algorithm.1 == "hierarchical") {

-      dist_mat <- dist(dataset, method = measure.1)

-      cl <- hclust(dist_mat, method = hier.agglo.algorithm.1)

+      dist_mat <- stats::dist(dataset, method = measure.1)

+      cl <- stats::hclust(dist_mat, method = hier.agglo.algorithm.1)

       temp <- data.frame(cutree(cl, k = i))

       cr1 <- cbind(cr1, temp)

     }

     # #k-means

     else if(algorithm.1 == "kmeans") {

-      cl <- kmeans(dataset, i, algorithm = measure.1)

+      cl <- stats::kmeans(dataset, i, algorithm = measure.1)

       temp <- data.frame(cl$cluster)

       cr1 <- cbind(cr1, temp)

     }

@@ -89,22 +92,22 @@ partitionAgreement <- function(data, algorithm.1 = "hierarchical",

   for(i in 2:number.of.clusters) {

     #Spectral clustering

     if(algorithm.2 == "spectral") {

-      cl <- specc(dataset, centers=i, kernel = measure.2)

+      cl <- kernlab::specc(dataset, centers=i, kernel = measure.2)

       temp <- data.frame([email protected])

       cr2 <- cbind(cr2, temp)

     }

     #Hierarchical

     else if(algorithm.2 == "hierarchical") {

-      dist_mat <- dist(dataset, method = measure.2)

-      cl <- hclust(dist_mat, method = hier.agglo.algorithm.2)

+      dist_mat <- stats::dist(dataset, method = measure.2)

+      cl <- stats::hclust(dist_mat, method = hier.agglo.algorithm.2)

       temp <- data.frame(cutree(cl, k = i))

       cr2 <- cbind(cr2, temp)

     }

     #k-means

     else if(algorithm.2 == "kmeans") {

-      cl <- kmeans(dataset, i, algorithm = measure.2)

+      cl <- stats::kmeans(dataset, i, algorithm = measure.2)

       temp <- data.frame(cl$cluster)

       cr2 <- cbind(cr2, temp)

     }

@@ -18,9 +18,10 @@ clusterVoting(data, min.k, max.k, algorithm)

 runs to be measured}

 }

 \value{

-An object of class "clusterVoting" containing a matrix with metric scores for every k and

-internal index, cluster memberships for every k, a dataframe with the k votes

- for every index, k vote frequencies and the frequency barplot of the k votes

+An object of class "clusterVoting" containing a matrix with metric

+ scores for every k and internal index, cluster memberships for every k, a

+ dataframe with the k votes for every index, k vote frequencies and the

+ frequency barplot of the k votes

 }

 \description{

 Estimating number of clusters through internal exhaustive ensemble majority

@@ -30,4 +31,5 @@ voting

 clusterVoting(toy_genes, 4,14,"sc")

 clusterVoting(toy_genes, 2,7,"hc")

 clusterVoting(toy_genes, 2,4,"km")

+

 }

@@ -27,4 +27,5 @@ clusteringMethodSelection(toy_genes, method.upper.k = 3,

 number.of.comparisons = 4)

 clusteringMethodSelection(toy_genes, method.upper.k = 2,

 number.of.comparisons = 2)

+

 }

@@ -28,4 +28,5 @@ clusters

 \examples{

 feasibilityAnalysis(classes = 3, samples = 320, features = 400)

 feasibilityAnalysis(classes = 4, samples = 400, features = 120)

+

 }

@@ -27,4 +27,5 @@ deviation

 \examples{

 feasibilityAnalysisDataBased(data = toy_genes, classes = 3)

 feasibilityAnalysisDataBased(data = toy_genes, classes = 2)

+

 }

@@ -23,4 +23,5 @@ Generating the feature/gene signature per cluster

 }

 \examples{

 geneSignatures(toy_genes, toy_gene_memberships)

+

 }

@@ -24,4 +24,5 @@ Clustering with the optimal parameters estimated by these tools

 \examples{

 optimalClustering(toy_genes, 4,"spectral")

 optimalClustering(toy_genes, 2,"kmeans")

+

 }