@@ -72,10 +72,10 @@ Collate:

     'rankingLimma.R'

     'rankingPairsDifferences.R'

     'rankingPlot.R'

-    'rankingSelectMulti.R'

     'runTest.R'

     'runTests.R'

     'samplesMetricMap.R'

+    'selectMulti.R'

     'selectionPlot.R'

     'simpleParams.R'

     'subtractFromLocation.R'

@@ -165,6 +165,7 @@ setMethod("crossValidate", "DataFrame",

                                   set.seed(seed)

                                   measurementsUse <- measurements

                                   if(!is.null(mcols(measurements))) measurementsUse <- measurements[, mcols(measurements)[, "dataset"] == dataIndex, drop = FALSE]

+                                  

                                   CV(

                                       measurements = measurementsUse, classes = classes,

                                       nFeatures = nFeatures[dataIndex],

@@ -233,7 +234,7 @@ setMethod("crossValidate", "DataFrame",

                   if(is.null(dataCombinations)){

-                      dataCombinations <- do.call("c", sapply(seq_len(length(datasetIDs)),function(n)combn(datasetIDs, n, simplify = FALSE)))

+                      dataCombinations <- do.call("c", sapply(seq_along(datasetIDs),function(n)combn(datasetIDs, n, simplify = FALSE)))

                       dataCombinations <- dataCombinations[sapply(dataCombinations, function(x)"clinical"%in%x, simplify = TRUE)]

                       if(length(dataCombinations)==0) stop("No dataCombinations with `clinical` data")

                   }

@@ -267,7 +268,7 @@ setMethod("crossValidate", "DataFrame",

                   if(is.null(dataCombinations)){

-                      dataCombinations <- do.call("c", sapply(seq_len(length(datasetIDs)),function(n)combn(datasetIDs, n, simplify = FALSE)))

+                      dataCombinations <- do.call("c", sapply(seq_along(datasetIDs),function(n)combn(datasetIDs, n, simplify = FALSE)))

                       dataCombinations <- dataCombinations[sapply(dataCombinations, function(x)"clinical"%in%x, simplify = TRUE)]

                       if(length(dataCombinations)==0) stop("No dataCombinations with `clinical` data")

                   }

@@ -611,8 +612,6 @@ generateModellingParams <- function(datasetIDs,

                                     classifier,

                                     multiViewMethod = "none"

 ){

-

-

     if(multiViewMethod != "none") {

         params <- generateMultiviewParams(datasetIDs,

                                           measurements,

@@ -627,7 +626,8 @@ generateModellingParams <- function(datasetIDs,

-    obsFeatures <- sum(mcols(measurements)[, "dataset"] %in% datasetIDs)

+    if(length(datasetIDs) > 1) obsFeatures <- sum(mcols(measurements)[, "dataset"] %in% datasetIDs)

+    else obsFeatures <- ncol(measurements)

     nFeatures <- unlist(nFeatures)

@@ -146,6 +146,7 @@ setMethod("SVMpredictInterface", c("svm", "DataFrame"), function(model, measurem

   # Prediction function depends on test data having same set of columns in same order as

   # selected features used for training.

+  colnames(measurementsTest) <- make.names(colnames(measurementsTest))

   measurementsTest <- measurementsTest[, colnames(model[["SV"]])]

   classPredictions <- predict(model, measurementsTest, probability = TRUE)

@@ -31,9 +31,8 @@

 #' @param verbose Default: 3. A number between 0 and 3 for the amount of

 #' progress messages to give.  This function only prints progress messages if

 #' the value is 3.

-#' @return A vector or data frame (if \code{MultiAssayExperiment} input) of

-#' features, from the most promising features in the first position to the

-#' least promising feature in the last position.

+#' @return A vector of feature indices, from the most promising features in

+#' the first position to the least promising feature in the last position.

 #' @author Dario Strbenac

 #' @examples

 #' 

@@ -29,9 +29,8 @@

 #' @param verbose Default: 3. A number between 0 and 3 for the amount of

 #' progress messages to give.  This function only prints progress messages if

 #' the value is 3.

-#' @return A vector or data frame (if \code{MultiAssayExperiment} input) of

-#' features, from the most promising features in the first position to the

-#' least promising feature in the last position.

+#' @return A vector of feature indicies, from the most promising features in the

+#' first position to the least promising feature in the last position.

 #' @importFrom survival coxph

 #' @rdname coxphRanking

 #' @usage NULL

@@ -32,9 +32,8 @@

 #' @param verbose Default: 3. A number between 0 and 3 for the amount of

 #' progress messages to give.  This function only prints progress messages if

 #' the value is 3.

-#' @return A vector or data frame (if \code{MultiAssayExperiment} input) of

-#' features, from the most promising features in the first position to the

-#' least promising feature in the last position.

+#' @return A vector of feature indices, from the most promising features in the

+#' first position to the least promising feature in the last position.

 #' @author Dario Strbenac

 #' @examples

 #' 

@@ -38,9 +38,8 @@

 #' @param verbose Default: 3. A number between 0 and 3 for the amount of

 #' progress messages to give.  This function only prints progress messages if

 #' the value is 3.

-#' @return A vector or data frame (if \code{MultiAssayExperiment} input) of

-#' features, from the most promising features in the first position to the

-#' least promising feature in the last position.

+#' @return A vector of feature indices, from the most promising features in the

+#' first position to the least promising feature in the last position.

 #' @author Dario Strbenac

 #' @references edgeR: a Bioconductor package for differential expression

 #' analysis of digital gene expression data, Mark D. Robinson, Davis McCarthy,

@@ -25,9 +25,8 @@

 #' @param verbose Default: 3. A number between 0 and 3 for the amount of

 #' progress messages to give.  This function only prints progress messages if

 #' the value is 3.

-#' @return A vector or data frame (if \code{MultiAssayExperiment} input) of

-#' features, from the most promising features in the first position to the

-#' least promising feature in the last position.

+#' @return A vector of feature indices, from the most promising features in the

+#' first position to the least promising feature in the last position.

 #' @author Dario Strbenac

 #' @examples

 #' 

@@ -34,9 +34,8 @@

 #' @param verbose Default: 3. A number between 0 and 3 for the amount of

 #' progress messages to give.  This function only prints progress messages if

 #' the value is 3.

-#' @return A vector or data frame (if \code{MultiAssayExperiment} input) of

-#' features, from the most promising features in the first position to the

-#' least promising feature in the last position.

+#' @return A vector of feature indices, from the most promising features in the

+#' first position to the least promising feature in the last position.

 #' @author Dario Strbenac

 #' @examples

 #' 

@@ -26,9 +26,8 @@

 #' @param verbose Default: 3. A number between 0 and 3 for the amount of

 #' progress messages to give.  This function only prints progress messages if

 #' the value is 3.

-#' @return A vector or data frame (if \code{MultiAssayExperiment} input) of

-#' features, from the most promising features in the first position to the

-#' least promising feature in the last position.

+#' @return A vector of feature indices, from the most promising features in the

+#' first position to the least promising feature in the last position.

 #' @author Dario Strbenac

 #' @examples

 #' 

@@ -37,9 +37,8 @@

 #' @param verbose Default: 3. A number between 0 and 3 for the amount of

 #' progress messages to give.  This function only prints progress messages if

 #' the value is 3.

-#' @return A vector or data frame (if \code{MultiAssayExperiment} input) of

-#' features, from the most promising features in the first position to the

-#' least promising feature in the last position.

+#' @return A vector of feature indices, from the most promising features in the

+#' first position to the least promising feature in the last position.

 #' @author Dario Strbenac

 #' @examples

 #' 

@@ -25,9 +25,8 @@

 #' @param verbose Default: 3. A number between 0 and 3 for the amount of

 #' progress messages to give.  This function only prints progress messages if

 #' the value is 3.

-#' @return A vector or data frame (if \code{MultiAssayExperiment} input) of

-#' features, from the most promising features in the first position to the

-#' least promising feature in the last position.

+#' @return A vector of feature indicies, from the most promising features in

+#' the first position to the least promising feature in the last position.

 #' @author Dario Strbenac

 #' @references Limma: linear models for microarray data, Gordon Smyth, 2005,

 #' In: Bioinformatics and Computational Biology Solutions using R and

@@ -31,7 +31,7 @@

 #' @param verbose Default: 3. A number between 0 and 3 for the amount of

 #' progress messages to give.  This function only prints progress messages if

 #' the value is 3.

-#' @return A \code{\link{Pairs}} object, from the most promising feature pair

+#' @return A vector of feature indices, from the most promising feature pair

 #' in the first position to the least promising feature pair in the last

 #' position.

 #' @author Dario Strbenac

@@ -119,9 +119,9 @@ function(measurementsTrain, outcomesTrain, measurementsTest, outcomesTest,

     {

       S4Vectors::mcols(measurementsTrain) <- featuresInfo[, c("Renamed Dataset", "Renamed Feature")]

       S4Vectors::mcols(measurementsTest) <- featuresInfo[, c("Renamed Dataset", "Renamed Feature")]

+      colnames(measurementsTrain) <- colnames(measurementsTest) <- paste(featuresInfo[["Renamed Dataset"]], featuresInfo[["Renamed Feature"]], sep = '')

     } else {

-      colnames(measurementsTrain) <- featuresInfo[, "Renamed Feature"]

-      colnames(measurementsTest) <- featuresInfo[, "Renamed Feature"]

+      colnames(measurementsTrain) <- colnames(measurementsTest) <- featuresInfo[, "Renamed Feature"]

     }

   }

@@ -185,8 +185,11 @@ input data. Autmomatically reducing to smaller number.")

     tuneDetailsSelect <- topFeatures[[3]]

     if(modellingParams@selectParams@subsetToSelections == TRUE)

+    {

       measurementsTrain <- measurementsTrain[, selectedFeaturesIndices, drop = FALSE]

-  } 

+      measurementsTest <- measurementsTest[, selectedFeaturesIndices, drop = FALSE]

+    }

+  }

   # Training stage.

   if(length(modellingParams@trainParams@intermediate) > 0)

@@ -232,25 +235,16 @@ input data. Autmomatically reducing to smaller number.")

   importanceTable <- NULL

   if(is.numeric(.iteration) && modellingParams@doImportance == TRUE)

   {

-    nSelected <- ifelse(is.null(ncol(selectedFeatures)), length(selectedFeatures), nrow(selectedFeatures))

     performanceMP <- modellingParams@selectParams@tuneParams[["performanceType"]]

     performanceType <- ifelse(!is.null(performanceMP), performanceMP, "Balanced Error")

-    performancesWithoutEach <- sapply(1:nSelected, function(selectedIndex)

+    performancesWithoutEach <- sapply(selectedFeaturesIndices, function(selectedIndex)

     {

-      if(is.null(S4Vectors::mcols(measurementsTrain)))

-      { # Input was ordinary matrix or DataFrame.

-        measurementsTrainLess1 <- measurementsTrain[, selectedFeatures[-selectedIndex], drop = FALSE]

-      } else { # Input was MultiAssayExperiment. # Match the selected features to the data frame columns

-        selectedIDs <-  do.call(paste, selectedFeatures[-selectedIndex, ])

-        featuresIDs <- do.call(paste, S4Vectors::mcols(measurementsTrain)[, c("dataset", "feature")])

-        useColumns <- match(selectedIDs, featuresIDs)

-        measurementsTrainLess1 <- measurementsTrain[, useColumns, drop = FALSE]

-      }

-         

-      modelWithoutOne <- tryCatch(.doTrain(measurementsTrainLess1, outcomesTrain, measurementsTest, outcomesTest, modellingParams, verbose),

+      measurementsTrainLess1 <- measurementsTrain[, -selectedIndex, drop = FALSE]

+      measurementsTestLess1 <- measurementsTest[, -selectedIndex, drop = FALSE]

+      modelWithoutOne <- tryCatch(.doTrain(measurementsTrainLess1, outcomesTrain, measurementsTestLess1, outcomesTest, modellingParams, verbose),

                                   error = function(error) error[["message"]])

       if(!is.null(modellingParams@predictParams))

-      predictedOutcomesWithoutOne <- tryCatch(.doTest(modelWithoutOne[["model"]], measurementsTest, modellingParams@predictParams, verbose),

+      predictedOutcomesWithoutOne <- tryCatch(.doTest(modelWithoutOne[["model"]], measurementsTestLess1, modellingParams@predictParams, verbose),

                                               error = function(error) error[["message"]])

       else predictedOutcomesWithoutOne <- modelWithoutOne[["model"]]

@@ -113,7 +113,7 @@ input data. Autmomatically reducing to smaller number.")

   {

     if(verbose >= 1 && setNumber %% 10 == 0)

       message("Processing sample set ", setNumber, '.')

-

+      

     # crossValParams is needed at least for nested feature tuning.

     runTest(measurements[trainingSamples, , drop = FALSE], outcomes[trainingSamples],

             measurements[testSamples, , drop = FALSE], outcomes[testSamples],

similarity index 64%

rename from R/rankingSelectMulti.R

rename to R/selectMulti.R

@@ -5,17 +5,15 @@ setGeneric("selectMulti", function(measurementsTrain, classesTrain, params, ...)

 setMethod("selectMulti", "DataFrame",

           function(measurementsTrain, classesTrain, params, verbose = 0)

           {

-              

-              assayTrain <- sapply(unique(mcols(measurementsTrain)[["dataset"]]), function(x) measurementsTrain[,mcols(measurementsTrain)[["dataset"]]%in%x], simplify = FALSE)

-              

-              selectedFeatures <- mapply(.doSelection, 

+              assayTrain <- sapply(unique(mcols(measurementsTrain)[["Renamed Dataset"]]), function(x) measurementsTrain[, mcols(measurementsTrain)[["Renamed Dataset"]] %in% x], simplify = FALSE)

+

+              featuresIndices <- mapply(.doSelection, 

                                          measurements = assayTrain,

-                                         modellingParams = params[names(assayTrain)],

+                                         modellingParams = params,

                                          MoreArgs = list(outcomesTrain = classesTrain, 

                                                          crossValParams = CrossValParams(permutations = 1, folds = 5), ###### Where to get this from?

-                                                         verbose = 0)

-              )

+                                                         verbose = 0), SIMPLIFY = FALSE

+                                        )

-              do.call("rbind", selectedFeatures[2,])

-              #S4Vectors::DataFrame(dataset = rep(names(selectedFeatures[2,]), unlist(lapply(selectedFeatures[2,], length))), feature = unlist(selectedFeatures[2,]))

+              unique(unlist(lapply(featuresIndices, "[[", 2)))

           })

@@ -354,25 +354,25 @@

       list(ranked = rankingUse, selected = selectionIndices, tune = tuneDetails)

     } else if(is.list(featureRanking)) { # It is a list of functions for ensemble selection.

-      featuresLists <- mapply(function(selector, selParams)

+      featuresIndiciesLists <- mapply(function(selector, selParams)

       {

         paramList <- list(measurementsTrain, outcomesTrain, trainParams = trainParams,

                           predictParams = predictParams, verbose = verbose)

         paramList <- append(paramList, selParams)

         do.call(selector, paramList)

-      }, modellingParams@selectParams@featureRanking, modellingParams@selectParams@featureRanking, SIMPLIFY = FALSE)

+      }, modellingParams@selectParams@featureRanking, modellingParams@selectParams@otherParams, SIMPLIFY = FALSE)

       performances <- sapply(topNfeatures, function(topN)

       {

-        topIndices <- unlist(lapply(featuresLists, function(features) features[1:topN]))

+        topIndices <- unlist(lapply(featuresIndiciesLists, function(featuresIndicies) featuresIndicies[1:topN]))

         topIndicesCounts <- table(topIndices)

         keep <- names(topIndicesCounts)[topIndicesCounts >= modellingParams@selectParams@minPresence]

-        measurementsSelected <- measurementsTrain[, keep, drop = FALSE] # Features in columns

+        measurementsTrain <- measurementsTrain[, as.numeric(keep), drop = FALSE] # Features in columns

         if(crossValParams@tuneMode == "Resubstitution")

         {

-          result <- runTest(measurementsSelected, classesTrain,

-                            training = 1:nrow(measurementsSelected), testing = 1:nrow(measurementsSelected),

+          result <- runTest(measurementsTrain, outcomesTrain,

+                            measurementsTrain, outcomesTrain,

                             crossValParams = NULL, modellingParams,

                             verbose = verbose, .iteration = "internal")

           predictions <- result[["predictions"]]

@@ -389,13 +389,13 @@

       })

       bestOne <- ifelse(betterValues == "lower", which.min(performances)[1], which.max(performances)[1])

-      selectedFeatures <- unlist(lapply(featuresLists, function(featuresList) featuresList[1:topNfeatures[bestOne]]))

-      names(table(selectedFeatures))[table(selectedFeatures) >= modellingParams@selectParams@minPresence]

+      selectionIndices <- unlist(lapply(featuresLists, function(featuresList) featuresList[1:topNfeatures[bestOne]]))

+      names(table(selectionIndices))[table(selectionIndices) >= modellingParams@selectParams@minPresence]

-      list(NULL, selectedFeatures, NULL)

+      list(NULL, selectionIndices, NULL)

     } else { # Previous selection

       selectedFeatures <- 

-      list(NULL, selectedFeatures, NULL)

+      list(NULL, selectionIndices, NULL)

     }

 }

@@ -45,9 +45,8 @@ and specifies that numeric variables from the sample information data table will

 used.}

 }

 \value{

-A vector or data frame (if \code{MultiAssayExperiment} input) of

-features, from the most promising features in the first position to the

-least promising feature in the last position.

+A vector of feature indices, from the most promising features in the

+first position to the least promising feature in the last position.

 }

 \description{

 Ranks features from largest Kolmogorov-Smirnov distance to smallest.

@@ -45,9 +45,8 @@ and specifies that numeric variables from the sample information data table will

 used.}

 }

 \value{

-A vector or data frame (if \code{MultiAssayExperiment} input) of

-features, from the most promising features in the first position to the

-least promising feature in the last position.

+A vector of feature indices, from the most promising features in the

+first position to the least promising feature in the last position.

 }

 \description{

 Ranks features from largest Kullback-Leibler distance between classes to

@@ -47,9 +47,8 @@ and specifies that numeric variables from the sample information table will be

 used.}

 }

 \value{

-A vector or data frame (if \code{MultiAssayExperiment} input) of

-features, from the most promising features in the first position to the

-least promising feature in the last position.

+A vector of feature indices, from the most promising features in

+the first position to the least promising feature in the last position.

 }

 \description{

 Ranks all features from largest Bartlett statistic to smallest.

@@ -45,9 +45,8 @@ and specifies that numeric variables from the clinical data table will be

 used.}

 }

 \value{

-A vector or data frame (if \code{MultiAssayExperiment} input) of

-features, from the most promising features in the first position to the

-least promising feature in the last position.

+A vector of feature indicies, from the most promising features in the

+first position to the least promising feature in the last position.

 }

 \description{

 Ranks all features from largest coxph statistic to smallest.

@@ -40,9 +40,8 @@ the value is 3.}

 used in the analysis.}

 }

 \value{

-A vector or data frame (if \code{MultiAssayExperiment} input) of

-features, from the most promising features in the first position to the

-least promising feature in the last position.

+A vector of feature indices, from the most promising features in the

+first position to the least promising feature in the last position.

 }

 \description{

 Uses an ordinary t-test if the data set has two classes or one-way ANOVA if

@@ -55,9 +55,8 @@ the value is 3.}

 names of the data tables of counts to be used.}

 }

 \value{

-A vector or data frame (if \code{MultiAssayExperiment} input) of

-features, from the most promising features in the first position to the

-least promising feature in the last position.

+A vector of feature indices, from the most promising features in the

+first position to the least promising feature in the last position.

 }

 \description{

 Performs a differential expression analysis between classes and ranks the

@@ -44,9 +44,8 @@ and specifies that numeric variables from the sample information table will be

 used.}

 }

 \value{

-A vector or data frame (if \code{MultiAssayExperiment} input) of

-features, from the most promising features in the first position to the

-least promising feature in the last position.

+A vector of feature indices, from the most promising features in the

+first position to the least promising feature in the last position.

 }

 \description{

 Ranks features by largest Levene statistic.

@@ -57,9 +57,8 @@ and specifies that numeric variables from the sample information data table will

 used.}

 }

 \value{

-A vector or data frame (if \code{MultiAssayExperiment} input) of

-features, from the most promising features in the first position to the

-least promising feature in the last position.

+A vector of feature indices, from the most promising features in the

+first position to the least promising feature in the last position.

 }

 \description{

 Ranks features from largest difference of log likelihoods (null hypothesis -

@@ -36,9 +36,8 @@ the value is 3.}

 used in the analysis.}

 }

 \value{

-A vector or data frame (if \code{MultiAssayExperiment} input) of

-features, from the most promising features in the first position to the

-least promising feature in the last position.

+A vector of feature indicies, from the most promising features in

+the first position to the least promising feature in the last position.

 }

 \description{

 Uses a moderated F-test with empirical Bayes shrinkage to rank

@@ -57,7 +57,7 @@ the value is 3.}

 name of the data table to be used.}

 }

 \value{

-A \code{\link{Pairs}} object, from the most promising feature pair

+A vector of feature indices, from the most promising feature pair

 in the first position to the least promising feature pair in the last

 position.

 }