diff --git a/DESCRIPTION b/DESCRIPTION
index 267fa1c5..3f1cd9e5 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -17,7 +17,6 @@ Depends:
     R (>= 3.6)
 Suggests: 
   curatedTCGAData,
-	TCGAutils,
 	rmarkdown,
 	testthat,
 	knitr,
diff --git a/NAMESPACE b/NAMESPACE
index 94091be0..98139214 100644
--- a/NAMESPACE
+++ b/NAMESPACE
@@ -15,6 +15,7 @@ export(convertProfileToNetworks)
 export(countIntType)
 export(countIntType_batch)
 export(countPatientsInNet)
+export(createNetFuncFromSimList)
 export(createPSN_MultiData)
 export(dataList2List)
 export(enrichLabelNets)
@@ -64,6 +65,7 @@ export(sparsify2)
 export(sparsify3)
 export(splitTestTrain)
 export(splitTestTrain_resampling)
+export(subsampleValidationData)
 export(tSNEPlotter)
 export(thresholdSmoothedMutations)
 export(updateNets)
diff --git a/R/buildPredictor.R b/R/buildPredictor.R
index a84fcebb..df05dad3 100644
--- a/R/buildPredictor.R
+++ b/R/buildPredictor.R
@@ -29,6 +29,9 @@
 #' So keys(groupList[["rna"]]) would have pathway names, generating one PSN
 #' per pathways, and values(groupList[["rna"]]) would be genes that would be
 #' grouped for the corresponding pathwayList.
+#' @param sims (list) rules to create similarity networks from input data. Keys are names of
+#' data layers and should be identical to names(groupList). Values is either a character
+#' for built-in similarity functions; call allowedSims() to see full list; or a custom function.
 #' @param makeNetFunc (function) user-defined function for creating the set
 #' of input PSN provided to netDx. See createPSN_MultiData()::customFunc.
 #' @param outDir (char) directory where results will be stored. If this
@@ -171,7 +174,8 @@
 #' #   makeNetFunc=makeNets, ### custom network creation function
 #' #   outDir=paste(tempdir(),"pred_output",sep=getFileSep()), ## absolute path
 #' #   numCores=16L,featScoreMax=2L, featSelCutoff=1L,numSplits=2L)
-buildPredictor <- function(dataList,groupList,outDir=tempdir(),makeNetFunc,
+buildPredictor <- function(dataList,groupList,outDir=tempdir(),
+	makeNetFunc=NULL,sims=NULL,
 	featScoreMax=10L,trainProp=0.8,numSplits=10L,numCores,JavaMemory=4L,
 	featSelCutoff=9L,keepAllData=FALSE,startAt=1L, preFilter=FALSE,
 	impute=FALSE,preFilterGroups=NULL, imputeGroups=NULL,logging="default",
@@ -196,7 +200,7 @@ if (logging == "all") {
 	verbose_predict <- FALSE
 }
 
-# Check input
+# Check input - error handling
 if (missing(dataList)) stop("dataList must be supplied.\n")
 if (missing(groupList)) stop("groupList must be supplied.\n")
 if (length(groupList)<1) stop("groupList must be of length 1+\n")
@@ -213,6 +217,9 @@ if (!is(groupList,"list") || not_list || names_nomatch ) {
 	stop(paste(msg,sep=""))
 }
 
+# checks either/or provided, sets missing var to NULL
+x <- checkMakeNetFuncSims(makeNetFunc=makeNetFunc, sims=sims,groupList=groupList)
+
 if (!is(dataList,"MultiAssayExperiment"))
 	stop("dataList must be a MultiAssayExperiment")
 
@@ -220,6 +227,8 @@ if (trainProp <= 0 | trainProp >= 1)
 		stop("trainProp must be greater than 0 and less than 1")
 if (startAt > numSplits) stop("startAt should be between 1 and numSplits")
 
+# end check input error handling
+
 megaDir <- outDir
 if (file.exists(megaDir)) {
  	stop(paste("outDir seems to already exist!",
@@ -274,7 +283,6 @@ if (verbose_default){
 	}
 }
 
-
 outList <- list()
 
 # create master list of possible networks
@@ -289,8 +297,14 @@ colnames(tmp) <- c("NetType","NetName")
 outList[["inputNets"]] <- tmp
 
 if (verbose_default) {
-	message("\n\nCustom function to generate input nets:")
-	print(makeNetFunc)
+	if (!is.null(makeNetFunc)){
+		message("\n\nCustom function to generate input nets:")
+		print(makeNetFunc)
+		
+	} else {
+		message("Similarity metrics provided:")
+		print(sims)
+	}
 	message(sprintf("-------------------------------\n"))
 }
 
@@ -386,8 +400,8 @@ for (rngNum in startAt:numSplits) {
 
 	if (verbose_default) message("** Creating features")
 	createPSN_MultiData(dataList=dats_train,groupList=groupList,
-			pheno=pheno_id,
-			netDir=netDir,customFunc=makeNetFunc,numCores=numCores,
+			pheno=pheno_id, 
+			netDir=netDir,makeNetFunc=makeNetFunc,sims=sims, numCores=numCores,
 			verbose=verbose_makeFeatures)
 	if (verbose_default) message("** Compiling features")
 	dbDir <- compileFeatures(netDir,outDir, numCores=numCores, 
@@ -512,7 +526,8 @@ for (rngNum in startAt:numSplits) {
 		pheno_id <- setupFeatureDB(pheno,netDir)
 		createPSN_MultiData(dataList=dats_tmp,groupList=groupList,
 			pheno=pheno_id,
-			netDir=netDir,customFunc=makeNetFunc,numCores=numCores,
+			netDir=netDir,makeNetFunc=makeNetFunc,sims=sims, 
+			numCores=numCores,
 			filterSet=pTally,verbose=verbose_default)
 		dbDir <- compileFeatures(netDir,outDir=pDir,numCores=numCores,
 			verbose=verbose_compileNets,debugMode=debugMode)
diff --git a/R/compileFeatures.R b/R/compileFeatures.R
index f5bb54b0..e7e7d4b8 100644
--- a/R/compileFeatures.R
+++ b/R/compileFeatures.R
@@ -58,7 +58,7 @@
 #' 
 #' pheno_id <- setupFeatureDB(pheno,netDir)
 #' netList <- createPSN_MultiData(dataList=dataList, groupList=groupList,
-#'     pheno=pheno_id,netDir=netDir,customFunc=makeNets,verbose=TRUE)
+#'     pheno=pheno_id,netDir=netDir,makeNetFunc=makeNets,verbose=TRUE)
 #' 
 #' outDir <- paste(tmpDir,'dbdir',sep=getFileSep()); 
 #'	dir.create(outDir)
diff --git a/R/createPSN_MultiData.R b/R/createPSN_MultiData.R
index ea6e976c..719497cd 100644
--- a/R/createPSN_MultiData.R
+++ b/R/createPSN_MultiData.R
@@ -12,11 +12,15 @@
 #' with internally-generated identifiers.
 #' @param netDir (char) path to directory where networks will be stored
 #' @param filterSet (char) vector of networks to include
-#' @param customFunc (function) custom user-function to create PSN. 
+#' @param makeNetFunc (function) custom user-function to create PSN. 
 #' Must take dataList,groupList,netDir as parameters. Must
 #' check if a given groupList is empty (no networks to create) before 
 #' the makePSN call for it. This is to avoid trying to make nets for datatypes
 #' that did not pass feature selection
+#' @param sims (list) Similarity metric settings for patient data. 
+#' Keys must be identical to those of groupList. 
+#' Values are either of type character, used for built-in similarity functions, 
+#' or are functions, when a custom function is provided.
 #' @param verbose (logical) print messages
 #' @param ... other parameters to makePSN_NamedMatrix() or makePSN_RangedSets()
 #' @return (char) vector of network names. Side effect of creating the nets
@@ -95,24 +99,24 @@
 #' pheno_id <- setupFeatureDB(colData(brca),netDir)
 #' createPSN_MultiData(dataList=datList2,groupList=groupList,
 #'  pheno=pheno_id,
-#'  netDir=netDir,customFunc=makeNets,numCores=1)
+#'  netDir=netDir,makeNetFunc=makeNets,numCores=1)
 #' @export
 createPSN_MultiData <- function(dataList, groupList, pheno, netDir=tempdir(), 
 		filterSet = NULL, 
-    verbose = TRUE, customFunc, ...) {
+    verbose = TRUE, makeNetFunc=NULL, sims=NULL, ...) {
     
     if (missing(dataList)) 
         stop("dataList must be supplied.\n")
     if (missing(groupList)) 
         stop("groupList must be supplied.\n")
-    
+ 
     # resolve user-provided IDs with internal IDs
     dataList <- lapply(dataList, function(x) {
         midx <- match(colnames(x), pheno$ID)
         colnames(x) <- pheno$INTERNAL_ID[midx]
         x
     })
-    
+
     if (!is.null(filterSet)) {
         if (length(filterSet) < 1) {
           s1 <- "filterSet is empty."
@@ -120,8 +124,8 @@ createPSN_MultiData <- function(dataList, groupList, pheno, netDir=tempdir(),
         	stop(paste(s1, s2, sep = " "))
 				}
     }
-    if (missing(customFunc)) 
-        stop("customFunc must be suppled.\n")
+
+    
     
     # Filter for nets (potentially feature-selected ones)
     if (!is.null(filterSet)) {
@@ -139,12 +143,22 @@ createPSN_MultiData <- function(dataList, groupList, pheno, netDir=tempdir(),
             }
         }
         groupList <- groupList2
+        sims <- sims[which(names(sims) %in% names(groupList))]
         rm(groupList2)
     }
     
+    if (!is.null(makeNetFunc)){
     # call user-defined function for making PSN
-    netList <- customFunc(dataList = dataList, groupList = groupList, 
+        netList <- makeNetFunc(dataList = dataList, groupList = groupList, 
 				netDir = netDir, ...)
+    } else {
+        netList <- createNetFuncFromSimList(dataList=dataList,
+            groupList = groupList, 
+            netDir = netDir,
+            sims = sims, 
+            ...
+            )
+    }
     
     if (length(netList) < 1) 
         stop("\n\nNo features created! Filters may be too stringent.\n")
diff --git a/R/helper.R b/R/helper.R
index e5e0bdec..02aa0e6c 100755
--- a/R/helper.R
+++ b/R/helper.R
@@ -159,8 +159,11 @@ return(list(
 #' @param EMapPctPass (numeric between 0 and 1) percent of splits for which feature must have score in range
 #'  [EMapMinScore,EMapMaxScore] to be included for EnrichmentMap visualization
 #' @param outDir (char) directory where files should be written
-#' @return 
-#' @export
+#' @return (list) 1) GMTfiles (char): GMT files used to create EnrichmentMap in Cytoscape.
+#' 2) NodeStyles (char): .txt files used to assign node attributes in Cytoscape. Importantly, 
+#' attributes include node fill, which indicates the highest consistent score for a given 
+#' feature. 
+#' @export 
 makeInputForEnrichmentMap <- function(model,results,pathwayList,
     EMapMinScore=0L, EMapMaxScore=1L,
     EMapPctPass=0.5,outDir)
@@ -214,8 +217,10 @@ return(list(GMTfiles=gmtFiles,NodeStyles=nodeAttrFiles))
 #' same class, relative to those of other classes, using Dijkstra distance (calcShortestPath flag).  
 #' @param dat (MultiAssayExperiment) input data
 #' @param groupList (list) feature groups, identical to groupList provided for buildPredictor()
-#' @param makeNets (function) Function used to create patient similarity networks. Identical to 
+#' @param makeNetFunc (function) Function used to create patient similarity networks. Identical to 
 #' makeNets provided to buildPredictor()
+#' @param sims (list) rules for creating PSN. Preferred over makeNetFunc. See buildPredictor() 
+#' for details.
 #' @param selectedFeatures (list) selected features for each class (key of list). This object is returned as
 #' part of a call to getResults(), after running buildPredictor().
 #' @param plotCytoscape (logical) If TRUE, plots network in Cytoscape.
@@ -245,9 +250,18 @@ return(list(GMTfiles=gmtFiles,NodeStyles=nodeAttrFiles))
 #' colours (colour)
 #' 6) outDir (char) value of outDir parameter
 #' @export
-getPSN <- function(dat, groupList, makeNets, selectedFeatures, plotCytoscape=FALSE,
-    aggFun="MEAN", prune_pctX=0.30, prune_useTop=TRUE,numCores=1L,calcShortestPath=FALSE
+getPSN <- function(dat, groupList, 
+    makeNetFunc=NULL, sims=NULL, 
+    selectedFeatures, plotCytoscape=FALSE,
+    aggFun="MEAN", prune_pctX=0.30, prune_useTop=TRUE,
+    numCores=1L,calcShortestPath=FALSE
     ){
+
+
+# checks either/or provided, sets missing var to NULL
+x <- checkMakeNetFuncSims(makeNetFunc=makeNetFunc, 
+    sims=sims,groupList=groupList)
+
 topPath <- gsub(".profile","", unique(unlist(selectedFeatures)))
 topPath <- gsub("_cont.txt","",topPath)
 
@@ -262,8 +276,10 @@ for (nm in names(groupList)) {
 
 message("* Making integrated PSN")
 psn <- 
-   plotIntegratedPatientNetwork(dat,
-  groupList=g2, makeNetFunc=makeNets,
+   plotIntegratedPatientNetwork(
+       dataList=dat,
+  groupList=g2, makeNetFunc=makeNetFunc,
+  sims=sims,
   aggFun=aggFun,
   prune_pctX=prune_pctX,
   prune_useTop=prune_useTop,
diff --git a/R/plotEmap.R b/R/plotEmap.R
index 2e1ff6bc..a15a994f 100644
--- a/R/plotEmap.R
+++ b/R/plotEmap.R
@@ -80,14 +80,14 @@ plotEmap <- function(gmtFile, nodeAttrFile, netName = "generic",
     }
     
     ####################################### create EM using given parameters
-    if (netName %in% getNetworkList()) {
+    if (netName %in% RCy3::getNetworkList()) {
         RCy3::deleteNetwork(netName)
     }
     em_command <- paste("enrichmentmap build analysisType=\"generic\"", 
 				"gmtFile=", gmtFile, "pvalue=", 1, "qvalue=", 1, 
 				"similaritycutoff=", 0.05, "coefficients=", "JACCARD")
     response <- RCy3::commandsGET(em_command)
-    renameNetwork(netName, getNetworkSuid())
+    RCy3::renameNetwork(netName, RCy3::getNetworkSuid())
     
     ### #annotate the network using AutoAnnotate app
     aa_command <- paste("autoannotate annotate-clusterBoosted", 
diff --git a/R/plotIntegratedPatientNetwork.R b/R/plotIntegratedPatientNetwork.R
index cd9ed4c4..4cb86750 100644
--- a/R/plotIntegratedPatientNetwork.R
+++ b/R/plotIntegratedPatientNetwork.R
@@ -19,6 +19,7 @@
 #' list of lists, where the outer list corresponds to assay (e.g. mRNA,
 #' clinical) and inner list to features to generate from that datatype.
 #' @param makeNetFunc (function) function to create features
+#' @param sims (list) rules for creating PSN. Preferred over makeNetFunc
 #' @param setName (char) name to assign the network in Cytoscape
 #' @param numCores (integer) number of cores for parallel processing
 #' @param prune_pctX (numeric between 0 and 1) fraction of most/least 
@@ -59,7 +60,8 @@
 #' @importFrom RColorBrewer brewer.pal
 #' @importFrom stats wilcox.test qexp density
 #' @export
-plotIntegratedPatientNetwork <- function(dataList,groupList,makeNetFunc,
+plotIntegratedPatientNetwork <- function(dataList,groupList,
+	makeNetFunc=NULL,sims=NULL,
 	setName="predictor",prune_pctX=0.05, prune_useTop=TRUE,
 	 aggFun="MAX",calcShortestPath=FALSE,
 	showStats=FALSE,
@@ -67,6 +69,11 @@ plotIntegratedPatientNetwork <- function(dataList,groupList,makeNetFunc,
 	nodeTransparency=155L,plotCytoscape=FALSE,
 	verbose=FALSE) {
 
+
+# checks either/or provided, sets missing var to NULL
+checkMakeNetFuncSims(makeNetFunc=makeNetFunc, 
+    sims=sims,groupList=groupList)
+
 if (missing(dataList)) stop("dataList is missing.")
 
 dat <- dataList2List(dataList, groupList)
@@ -81,7 +88,9 @@ pheno_id <- setupFeatureDB(pheno,outDir)
 
 createPSN_MultiData(dataList=dat$assays,groupList=groupList,
 			pheno=pheno_id,
-			netDir=outDir,customFunc=makeNetFunc,numCores=numCores,
+			netDir=outDir,
+			makeNetFunc=makeNetFunc,sims=sims,
+			numCores=numCores,
 			verbose=FALSE)
 convertProfileToNetworks(
 		netDir=profDir,
diff --git a/R/predict.R b/R/predict.R
index 4fd8ee0a..0df38814 100644
--- a/R/predict.R
+++ b/R/predict.R
@@ -5,12 +5,12 @@
 #' @param testMAE (MultiAssayExperiment) new patient dataset for testing model. Assays must be the same as for trainMAE.
 #' @param groupList (list) list of features used to train the model. Keys are data types, and values are lists for groupings within those datatypes.
 #' e.g. keys could include {'clinical','rna','methylation'}, and values within 'rna' could include pathway names {'cell cycle', 'DNA repair'}, etc.,
-#' featSel will be used to subset 
-#' @param featSel (list) selected features to be used in the predictive model. 
+#' selectedFeatures will be used to subset 
+#' @param selectedFeatures (list) selected features to be used in the predictive model. 
 #' keys are patient labels (e.g. "responder/nonresponder"), and values are feature names 
 #' identified by running buildPredictor(). Feature names must correspond to names of groupList, from which they will be subset.
 #' @param makeNetFunc (function) function to create PSN features from patient data. See makeNetFunc in buildPredictor() for details
-#' @param impute (logical) if TRUE imputes train and test samples separately before creating features. Currently unsupported.
+#' @param sims (list) rules for creating PSN. Preferred over makeNetFunc.
 #' @param outDir (char) directory for results
 #' @param verbose (logical) print messages
 #' @param numCores (integer) number of CPU cores for parallel processing
@@ -20,19 +20,20 @@
 #' columns are:  1) ID, 2) STATUS (ground truth), 3) <label>_SCORE: similarity score for the corresponding label,
 #' 4) PRED_CLASS: predicted class
 #' @export
-predict <- function(trainMAE, testMAE, groupList, featSel, makeNetFunc, outDir,
-    impute = FALSE, verbose = FALSE, numCores = 1L, JavaMemory = 4L, debugMode = FALSE) {
+predict <- function(trainMAE, testMAE, groupList, 
+  selectedFeatures, 
+  makeNetFunc=NULL, sims=NULL,
+  outDir, verbose = FALSE, 
+  numCores = 1L, JavaMemory = 4L, debugMode = FALSE) {
 
   # input checks
   if (missing(trainMAE)) stop("trainMAE must be supplied.\n")
   if (missing(testMAE)) stop("testMAE must be supplied.\n")
   if (missing(groupList)) stop("groupList must be supplied.\n")
   if (length(groupList) < 1) stop("groupList must be of length 1+\n")
-  if (class(featSel) != "list") stop("featSel must be a list with patient labels as keys, and selected features as values")
+  if (class(selectedFeatures) != "list") stop("selectedFeatures must be a list with patient labels as keys, and selected features as values")
   if (missing(outDir)) stop("outDir must be supplied.\n")
-  if (impute) stop("impute=TRUE is not supported in the current version of netDx. This will be implemented based on future user interest. Please contact Shraddha Pai <shraddha.pai@utoronto.ca> if this feature is required.")
 
-  nm1 <- setdiff(names(groupList), "clinical")
   if (!is(trainMAE, "MultiAssayExperiment"))
     stop("trainMAE must be a MultiAssayExperiment")
   if (!is(testMAE, "MultiAssayExperiment"))
@@ -41,6 +42,7 @@ predict <- function(trainMAE, testMAE, groupList, featSel, makeNetFunc, outDir,
   tmp <- unlist(lapply(groupList, class))
   not_list <- sum(tmp == "list") < length(tmp)
 
+  nm1 <- setdiff(names(groupList), "clinical")
   names_nomatch <- any(!nm1 %in% names(trainMAE))
   if (!is(groupList, "list") || not_list || names_nomatch) {
     msg <- c("groupList must be a list of lists.",
@@ -49,11 +51,13 @@ predict <- function(trainMAE, testMAE, groupList, featSel, makeNetFunc, outDir,
     stop(paste(msg, sep = ""))
   }
 
-  for (nm in names(featSel)) {
-    featSel[[nm]] <- sub("_cont.txt", "", sub(".profile", "", featSel[[nm]]))
+  for (nm in names(selectedFeatures)) {
+    selectedFeatures[[nm]] <- sub(
+      "_cont.txt", "", 
+      sub(".profile", "", selectedFeatures[[nm]]))
   }
   # clean features
-  fs <- unlist(featSel);
+  fs <- unlist(selectedFeatures);
   names(fs) <- NULL
   gl <- c()
   for (k in names(groupList)) {
@@ -62,7 +66,7 @@ predict <- function(trainMAE, testMAE, groupList, featSel, makeNetFunc, outDir,
   }
 
   if (sum(!fs %in% gl) > 0) {
-    stop("One or more entry in featSelNet not found in groupList.")
+    stop("One or more entry in selectedFeatures was not found in groupList.")
   }
 
   # merging train-test for joint db
@@ -99,24 +103,38 @@ predict <- function(trainMAE, testMAE, groupList, featSel, makeNetFunc, outDir,
   message("* Measuring similarity to each known class")
   subtypes <- unique(ph$STATUS)
   predRes <- list()
+
+  # classification for a given class is performed as follows:
+  # you take just the selected features for that class and create a single PSN comprised of the union
+  # of training and test samples.
+  # using training sampless for that class ("samples look like non-responders",e.g.) run label propagation
+  # on the PSN. get a similarity ranking for all test samples
+  # now repeat this exercise for all classes.
+  # ultimately, the patient is classified as the class to which they have greatest similarity.
   for (g in subtypes) {
     if (verbose) message(sprintf("\t%s", g))
     pDir <- paste(outDir, g, sep = getFileSep())
-    dir.create(pDir)
-
     netDir <- paste(pDir, "networks", sep = getFileSep())
+    dir.create(pDir)
     dir.create(netDir)
+
+    # make nets using only features selected for the label in question
     pheno_id <- setupFeatureDB(pheno, netDir)
 
+    # checks either/or provided, sets missing var to NULL
+    x <- checkMakeNetFuncSims(makeNetFunc=makeNetFunc, sims=sims,groupList=groupList)
+
     if (verbose) message("Creating PSN")
     createPSN_MultiData(dataList = assays,
         groupList = groupList,
         pheno = pheno_id,
         netDir = netDir,
-        customFunc = makeNetFunc,
+        makeNetFunc = makeNetFunc,
+        sims = sims,
         numCores = 1L,
-        filterSet = featSel[[g]],
+        filterSet = selectedFeatures[[g]],
         verbose = verbose)
+
     dbDir <- compileFeatures(netDir,
       outDir = pDir,
       numCores = numCores,
@@ -126,9 +144,10 @@ predict <- function(trainMAE, testMAE, groupList, featSel, makeNetFunc, outDir,
     # run query for this class
     qSamps <- pheno$ID[which(pheno$STATUS %in% g & pheno$TT_STATUS %in% "TRAIN")]
     qFile <- paste(pDir, sprintf("%s_query", g), sep = getFileSep())
-
     message(sprintf("\t%s : %s training samples", g, prettyNum(length(qSamps), big.mark = ",")))
     writeQueryFile(qSamps, "all", nrow(pheno), qFile)
+
+
     if (verbose) message(sprintf("\t** %s: Compute similarity", g))
     resFile <- runQuery(dbDir$dbDir, qFile, resDir = pDir,
       JavaMemory = JavaMemory, numCores = numCores,
@@ -136,9 +155,18 @@ predict <- function(trainMAE, testMAE, groupList, featSel, makeNetFunc, outDir,
     predRes[[g]] <- getPatientRankings(sprintf("%s.PRANK", resFile), pheno, g)
   }
 
+  # at this point, we should have similarity rankings for each of the test patients, for each of the classes.
   predClass <- predictPatientLabels(predRes,
       verbose = verbose)
   out <- merge(x = pheno, y = predClass, by = "ID")
+  if (nrow(out)!= nrow(colData(testMAE))) {
+      warning(
+        paste(rep("*",25),
+          "Not all patients provided in the test sample were classified.",
+              rep("*",25), sep="\n")
+      )
+  }
+
   acc <- sum(out$STATUS == out$PRED_CLASS) / nrow(out)
   message(sprintf("%s test patients", prettyNum(nrow(out), big.mark = ",")))
   message(sprintf("ACCURACY (N=%i test) = %2.1f%%",
diff --git a/R/similarities.R b/R/similarities.R
index b4c988e0..a377c911 100644
--- a/R/similarities.R
+++ b/R/similarities.R
@@ -206,3 +206,190 @@ avgNormDiff <- function(x) {
     sim <- sim/nrow(x)
     sim
 }
+
+#' built-in similarity functions
+#'
+allowedSims <- function(){
+  return(c("pearsonCorr","normDiff","avgNormDiff",
+        "sim.pearscale","sim.eucscale"))
+}
+
+
+
+#' checks if provided similarity functions are valid. Returns error if not
+#'
+#' @param sims (list) keys are layer names, values are functions or characters (names of built-in similarity functions)
+#' @return TRUE if all pass check. Else throws error.
+checkSimValid <- function(sims){
+    allowed <- allowedSims()
+    for (k in names(sims)){    
+        if (class(sims[[k]])!="function"){
+            if (class(sims[[k]])!="character"){
+                stop(paste("Invalid sims datatype. ",
+                    "sims entries must be functions or keywords (characters) ",
+                    "for built-in similarity functions.",sep=""))
+            } else {
+                if (!sims[[k]] %in% allowed){
+                    stop(paste(
+                            sprintf("sims[[%s]] has invalid similarity type:",k),
+                            sims[[k]],". ",
+                            "Allowed values are: {%s}",
+                            paste(allowed,collapse=",")))
+                }
+            }
+        }
+    }
+    return(TRUE)
+}
+
+#' internal test function to check validity of makeNetFunc and sims
+#'
+#' @details User must provide either makeNetFunc or sims. This function
+#' confirms this.
+#' @param makeNetFunc (function) makeNetFunc from buildPredictor()
+#' @param sims (list) sims from buildPredictor()
+#' @param groupList (list) groupList from buildPredictor()s
+#' @return (list) cleaned values for makeNetFunc and Sims
+checkMakeNetFuncSims <- function(makeNetFunc,sims,groupList){
+    if (is.null(makeNetFunc) && is.null(sims)) {
+	stop("Provide either makeNetFunc or sims (preferred).")
+} 
+if (!is.null(makeNetFunc) && !is.null(sims)){
+	stop("Provide either makeNetFunc or sims (preferred).")
+}
+
+if (!is.null(sims))	{
+	if (class(sims)!="list") stop("sims must be a list.")
+	if (all.equal(sort(names(sims)),sort(names(groupList)))!=TRUE) 
+		stop("names(sims) must match names(groupList).")
+}
+return(TRUE)
+}
+
+#' Create PSN from provided similarities
+#'
+#' @details Called by CreatePSN_MultiData(), this is the function that converts user-provided
+#' simlarity metrics to internal netDx function calls to generate nets.
+#' @param dataList (list) patient data, output of dataList2List()
+#' @param groupList (list) measure groupings. Keys match assays(dataList) and are usually different data sources. Values for each are a list of 
+#' networks with user-provided groupings. See groupList in buildPredictor() for details.
+#' @param netDir (char) path to directory where networks are to be created
+#' @param sims (list) keys must be identical to those of groupList. Values are either of type character, used for built-in similarity functions, 
+#' or are functions, when a custom function is provided.
+#' @param verbose (logical) print messages
+#' @param ... values to be passed to PSN creation functions such as makePSN_NamedMatrix().
+#' @export
+createNetFuncFromSimList <- function(dataList, groupList, netDir, sims,
+    verbose=TRUE,...){    
+    
+    if (length(groupList)!= length(sims)){
+        stop("groupList and sims need to be of same length.")
+    }
+    if (all.equal(sort(names(groupList)),sort(names(sims)))!=TRUE){
+        stop("names(groupList) needs to match names(sims).")
+    }
+    settings <- list(dataList=dataList,groupList=groupList,
+                    netDir=netDir,sims=sims)
+
+    if (verbose) message("Making nets from sims")
+    netList <- c()    
+    for (nm in names(sims)){
+        csim <- sims[[nm]]
+        netList_cur <- NULL
+        if (verbose) message(sprintf("\t%s",nm))
+
+        cur_set <- settings; 
+        cur_set[["name"]] <- nm; cur_set[["similarity"]] <- csim
+
+        if (!is.null(groupList[[nm]])){
+            if (class(csim)=="function") {# custom function
+    
+                netList_cur <- psn__custom(cur_set,csim, verbose,...)
+            } else if (csim == "pearsonCorr") {
+                netList_cur <- psn__corr(cur_set,verbose,...)
+            } else {
+                netList_cur <- psn__builtIn(cur_set,verbose,...)
+            }
+            netList <- c(netList,netList_cur)
+        }
+    }
+    if (verbose) {
+        message("Net construction complete!")
+    }
+    unlist(netList)
+}
+
+#' make PSN for built-in similarity functions
+#'
+#' @param settings (list) from makeNetFunc
+#' @param verbose (logical) print messages
+#' @param ... parameters for makePSN_NamedMatrix()
+#' @return (char) names of networks created. Side effect of network creation.
+psn__builtIn <- function(settings,verbose,...){
+
+funcs <- list(
+    "normDiff"=normDiff,
+    "avgNormDiff"=avgNormDiff,
+    "sim.pearscale"=sim.pearscale,
+    "sim.eucscale"=sim.eucscale
+)
+
+    if (verbose) message(sprintf("Layer %s: Built-in function %s",
+            settings$name,settings$similarity))
+
+    nm <- settings$name
+    netList <- makePSN_NamedMatrix(
+        settings$dataList[[nm]],
+		rownames(settings$dataList[[nm]]),
+		settings$groupList[[nm]],
+        settings$netDir,
+		simMetric="custom",
+        customFunc=funcs[[settings$similarity]], # custom function
+		writeProfiles=FALSE,
+		sparsify=TRUE,...
+    )
+    netList
+}
+
+#' make PSN for custom similarity functions
+#'
+#' @param settings (list) from makeNetFunc
+#' @param fn (function) custom similarity function
+#' @param verbose (logical) print messages
+#' @param ... parameters for makePSN_NamedMatrix()
+#' @return (char) names of networks created. Side effect of network creation.
+psn__custom <- function(settings,fn,verbose, ...){
+    nm <- settings$name
+    if (verbose) message(sprintf("Layer %s: CUSTOM FUNCTION",settings$name))
+    netList <- makePSN_NamedMatrix(
+        settings$dataList[[nm]],
+		rownames(settings$dataList[[nm]]),
+		settings$groupList[[nm]],
+        settings$netDir,
+		simMetric="custom",customFunc=fn, # custom function
+		writeProfiles=FALSE,
+		sparsify=TRUE,...
+    )
+    netList
+}
+
+#' wrapper for PSNs using Pearson correlation
+#'
+#' @param settings (list) from makeNetFunc
+#' @param verbose (logical) print messages
+#' @param ... parameters for makePSN_NamedMatrix()
+#' @return (char) names of networks created. Side effect of network creation.
+psn__corr <- function(settings,verbose,...){
+    if (verbose) message(sprintf("Layer %s: PEARSON CORR",settings$name))
+    nm <- settings$name
+    netList <- makePSN_NamedMatrix(
+				xpr=settings$dataList[[nm]],
+				nm=rownames(settings$dataList[[nm]]),
+				namedSets=settings$groupList[[nm]],	
+				outDir=settings$netDir,	
+				verbose=FALSE, 			
+				writeProfiles=TRUE,  
+				...
+				)
+    return(netList)
+}
\ No newline at end of file
diff --git a/R/splitTestTrain_partition.R b/R/splitTestTrain_partition.R
index b14f9ab2..667b03d7 100644
--- a/R/splitTestTrain_partition.R
+++ b/R/splitTestTrain_partition.R
@@ -79,3 +79,44 @@ splitTestTrain_resampling <- function(pheno_DF, nFold = 3L, predClass,
     
     out
 }
+
+
+#' Subsample a hold-out set from a larger patient dataset
+#'
+#' @details Creates a partition of data to be used for model validation after initial model building.
+#' In netDx, buildPredictor() is used for model training, and selected features from this exercise
+#' are used to validate a held-out dataset with the predict() function. Note that this function
+#' identifies a random subsample, which may result in a validation sample that is not representative
+#' of your training bias. If this method is used, please use data exploration techniques (e.g. UMAP)
+#' to ensure that validation accuracy is not confounded by stratification.
+#' @param dataMAE (MultiAssayExperiment) patient data to be subsampled. Must have columns ID (patient ID)
+#' and STATUS
+#' @param pctValidation (numeric) Fraction of dataset to include in the validation set. Value from 0.05 to 0.95.
+#' @param verbose (logical)
+#' @return (list) Keys are trainMAE and validationMAE. These contain corresponding MultiAssayExperiments for
+#' training and test data
+#' @export
+subsampleValidationData <- function(dataMAE,pctValidation=0.2,verbose=TRUE) {
+    if (pctValidation < 0.05 || pctValidation > 0.95) stop("pctValidation should be between 0.05 and 0.95.")
+    if (missing(dataMAE)) stop("Supply dataMAE.")
+    if (class(dataMAE)!= "MultiAssayExperiment") stop("dataMAE must be an object of type MultiAssayExperiment.")
+
+    pheno <- colData(dataMAE)
+    st <- unique(pheno$STATUS)
+    nsamp <- round(pctValidation/length(st) * nrow(pheno))
+
+    idx_holdout <- c()
+    for (k in unique(pheno$STATUS)) {
+        idx_holdout <- c(idx_holdout, 
+            sample(which(pheno$STATUS == k),nsamp,FALSE)
+        )  
+    }
+
+    holdout <- dataMAE[,rownames(pheno)[idx_holdout]]
+    colData(holdout)$ID <- as.character(colData(holdout)$ID)
+    tokeep <- setdiff(1:nrow(pheno),idx_holdout)
+
+    dataMAE <- dataMAE[,rownames(pheno)[tokeep]]
+
+    return(list(trainMAE=dataMAE, validationMAE=holdout))
+}
\ No newline at end of file
diff --git a/data/cnv_GR.rda b/data/cnv_GR.rda
index e4245fad..24ee75aa 100644
Binary files a/data/cnv_GR.rda and b/data/cnv_GR.rda differ
diff --git a/data/cnv_patientNetCount.rda b/data/cnv_patientNetCount.rda
index 0a72e9d3..a5739127 100644
Binary files a/data/cnv_patientNetCount.rda and b/data/cnv_patientNetCount.rda differ
diff --git a/data/cnv_pheno.rda b/data/cnv_pheno.rda
index 2431f8ca..ba3fbc31 100644
Binary files a/data/cnv_pheno.rda and b/data/cnv_pheno.rda differ
diff --git a/data/toymodel.rda b/data/toymodel.rda
index caa1293c..e9c92383 100644
Binary files a/data/toymodel.rda and b/data/toymodel.rda differ
diff --git a/data/xpr.rda b/data/xpr.rda
index 37353053..aa7c22bc 100644
Binary files a/data/xpr.rda and b/data/xpr.rda differ
diff --git a/man/allowedSims.Rd b/man/allowedSims.Rd
new file mode 100644
index 00000000..06b08570
--- /dev/null
+++ b/man/allowedSims.Rd
@@ -0,0 +1,11 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/similarities.R
+\name{allowedSims}
+\alias{allowedSims}
+\title{built-in similarity functions}
+\usage{
+allowedSims()
+}
+\description{
+built-in similarity functions
+}
diff --git a/man/buildPredictor.Rd b/man/buildPredictor.Rd
index b0480695..df8435c0 100644
--- a/man/buildPredictor.Rd
+++ b/man/buildPredictor.Rd
@@ -8,7 +8,8 @@ buildPredictor(
   dataList,
   groupList,
   outDir = tempdir(),
-  makeNetFunc,
+  makeNetFunc = NULL,
+  sims = NULL,
   featScoreMax = 10L,
   trainProp = 0.8,
   numSplits = 10L,
@@ -49,6 +50,10 @@ directory exists, its contents will be overwritten. Must be absolute path}
 \item{makeNetFunc}{(function) user-defined function for creating the set
 of input PSN provided to netDx. See createPSN_MultiData()::customFunc.}
 
+\item{sims}{(list) rules to create similarity networks from input data. Keys are names of
+data layers and should be identical to names(groupList). Values is either a character
+for built-in similarity functions; call allowedSims() to see full list; or a custom function.}
+
 \item{featScoreMax}{(integer) number of CV folds in inner loop}
 
 \item{trainProp}{(numeric 0 to 1) Percent samples to use for training}
diff --git a/man/buildPredictor_sparseGenetic.Rd b/man/buildPredictor_sparseGenetic.Rd
index 7dbfce64..e8729bbb 100644
--- a/man/buildPredictor_sparseGenetic.Rd
+++ b/man/buildPredictor_sparseGenetic.Rd
@@ -147,7 +147,7 @@ cnv_GR    <- GRanges(pheno$seqnames,IRanges(pheno$start,pheno$end),
                         ID=pheno$ID,LOCUS_NAMES=pheno$Gene_symbols)
 
 # get gene coordinates
-geneURL <- paste("https://download.baderlab.org/netDx/",
+geneURL <- paste("http://download.baderlab.org/netDx/",
 	"supporting_data/refGene.hg18.bed",sep="")
 cache <- rappdirs::user_cache_dir(appname = "netDx")
 bfc <- BiocFileCache::BiocFileCache(cache,ask=FALSE)
diff --git a/man/checkMakeNetFuncSims.Rd b/man/checkMakeNetFuncSims.Rd
new file mode 100644
index 00000000..3c8a7dbf
--- /dev/null
+++ b/man/checkMakeNetFuncSims.Rd
@@ -0,0 +1,25 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/similarities.R
+\name{checkMakeNetFuncSims}
+\alias{checkMakeNetFuncSims}
+\title{internal test function to check validity of makeNetFunc and sims}
+\usage{
+checkMakeNetFuncSims(makeNetFunc, sims, groupList)
+}
+\arguments{
+\item{makeNetFunc}{(function) makeNetFunc from buildPredictor()}
+
+\item{sims}{(list) sims from buildPredictor()}
+
+\item{groupList}{(list) groupList from buildPredictor()s}
+}
+\value{
+(list) cleaned values for makeNetFunc and Sims
+}
+\description{
+internal test function to check validity of makeNetFunc and sims
+}
+\details{
+User must provide either makeNetFunc or sims. This function
+confirms this.
+}
diff --git a/man/checkSimValid.Rd b/man/checkSimValid.Rd
new file mode 100644
index 00000000..ca6626e2
--- /dev/null
+++ b/man/checkSimValid.Rd
@@ -0,0 +1,17 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/similarities.R
+\name{checkSimValid}
+\alias{checkSimValid}
+\title{checks if provided similarity functions are valid. Returns error if not}
+\usage{
+checkSimValid(sims)
+}
+\arguments{
+\item{sims}{(list) keys are layer names, values are functions or characters (names of built-in similarity functions)}
+}
+\value{
+TRUE if all pass check. Else throws error.
+}
+\description{
+checks if provided similarity functions are valid. Returns error if not
+}
diff --git a/man/compileFeatures.Rd b/man/compileFeatures.Rd
index 4b79eb32..c4b85cb4 100644
--- a/man/compileFeatures.Rd
+++ b/man/compileFeatures.Rd
@@ -97,7 +97,7 @@ dir.create(netDir,recursive=TRUE)
 
 pheno_id <- setupFeatureDB(pheno,netDir)
 netList <- createPSN_MultiData(dataList=dataList, groupList=groupList,
-    pheno=pheno_id,netDir=netDir,customFunc=makeNets,verbose=TRUE)
+    pheno=pheno_id,netDir=netDir,makeNetFunc=makeNets,verbose=TRUE)
 
 outDir <- paste(tmpDir,'dbdir',sep=getFileSep()); 
 dir.create(outDir)
diff --git a/man/createNetFuncFromSimList.Rd b/man/createNetFuncFromSimList.Rd
new file mode 100644
index 00000000..82a5142f
--- /dev/null
+++ b/man/createNetFuncFromSimList.Rd
@@ -0,0 +1,37 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/similarities.R
+\name{createNetFuncFromSimList}
+\alias{createNetFuncFromSimList}
+\title{Create PSN from provided similarities}
+\usage{
+createNetFuncFromSimList(
+  dataList,
+  groupList,
+  netDir,
+  sims,
+  verbose = TRUE,
+  ...
+)
+}
+\arguments{
+\item{dataList}{(list) patient data, output of dataList2List()}
+
+\item{groupList}{(list) measure groupings. Keys match assays(dataList) and are usually different data sources. Values for each are a list of 
+networks with user-provided groupings. See groupList in buildPredictor() for details.}
+
+\item{netDir}{(char) path to directory where networks are to be created}
+
+\item{sims}{(list) keys must be identical to those of groupList. Values are either of type character, used for built-in similarity functions, 
+or are functions, when a custom function is provided.}
+
+\item{verbose}{(logical) print messages}
+
+\item{...}{values to be passed to PSN creation functions such as makePSN_NamedMatrix().}
+}
+\description{
+Create PSN from provided similarities
+}
+\details{
+Called by CreatePSN_MultiData(), this is the function that converts user-provided
+simlarity metrics to internal netDx function calls to generate nets.
+}
diff --git a/man/createPSN_MultiData.Rd b/man/createPSN_MultiData.Rd
index 3e6d45e5..200be81a 100644
--- a/man/createPSN_MultiData.Rd
+++ b/man/createPSN_MultiData.Rd
@@ -11,7 +11,8 @@ createPSN_MultiData(
   netDir = tempdir(),
   filterSet = NULL,
   verbose = TRUE,
-  customFunc,
+  makeNetFunc = NULL,
+  sims = NULL,
   ...
 )
 }
@@ -35,12 +36,17 @@ with internally-generated identifiers.}
 
 \item{verbose}{(logical) print messages}
 
-\item{customFunc}{(function) custom user-function to create PSN. 
+\item{makeNetFunc}{(function) custom user-function to create PSN. 
 Must take dataList,groupList,netDir as parameters. Must
 check if a given groupList is empty (no networks to create) before 
 the makePSN call for it. This is to avoid trying to make nets for datatypes
 that did not pass feature selection}
 
+\item{sims}{(list) Similarity metric settings for patient data. 
+Keys must be identical to those of groupList. 
+Values are either of type character, used for built-in similarity functions, 
+or are functions, when a custom function is provided.}
+
 \item{...}{other parameters to makePSN_NamedMatrix() or makePSN_RangedSets()}
 }
 \value{
@@ -124,5 +130,5 @@ netDir <- tempdir()
 pheno_id <- setupFeatureDB(colData(brca),netDir)
 createPSN_MultiData(dataList=datList2,groupList=groupList,
  pheno=pheno_id,
- netDir=netDir,customFunc=makeNets,numCores=1)
+ netDir=netDir,makeNetFunc=makeNets,numCores=1)
 }
diff --git a/man/getPSN.Rd b/man/getPSN.Rd
index b7c342e5..1e37f671 100644
--- a/man/getPSN.Rd
+++ b/man/getPSN.Rd
@@ -7,7 +7,8 @@
 getPSN(
   dat,
   groupList,
-  makeNets,
+  makeNetFunc = NULL,
+  sims = NULL,
   selectedFeatures,
   plotCytoscape = FALSE,
   aggFun = "MEAN",
@@ -22,9 +23,12 @@ getPSN(
 
 \item{groupList}{(list) feature groups, identical to groupList provided for buildPredictor()}
 
-\item{makeNets}{(function) Function used to create patient similarity networks. Identical to 
+\item{makeNetFunc}{(function) Function used to create patient similarity networks. Identical to 
 makeNets provided to buildPredictor()}
 
+\item{sims}{(list) rules for creating PSN. Preferred over makeNetFunc. See buildPredictor() 
+for details.}
+
 \item{selectedFeatures}{(list) selected features for each class (key of list). This object is returned as
 part of a call to getResults(), after running buildPredictor().}
 
diff --git a/man/makeInputForEnrichmentMap.Rd b/man/makeInputForEnrichmentMap.Rd
index 07f7ef89..fab88020 100644
--- a/man/makeInputForEnrichmentMap.Rd
+++ b/man/makeInputForEnrichmentMap.Rd
@@ -31,7 +31,10 @@ makeInputForEnrichmentMap(
 \item{outDir}{(char) directory where files should be written}
 }
 \value{
-
+(list) 1) GMTfiles (char): GMT files used to create EnrichmentMap in Cytoscape.
+2) NodeStyles (char): .txt files used to assign node attributes in Cytoscape. Importantly, 
+attributes include node fill, which indicates the highest consistent score for a given 
+feature.
 }
 \description{
 Wrapper to create input files for Enrichment Map
diff --git a/man/plotIntegratedPatientNetwork.Rd b/man/plotIntegratedPatientNetwork.Rd
index 3166a8b9..5ef12fc6 100644
--- a/man/plotIntegratedPatientNetwork.Rd
+++ b/man/plotIntegratedPatientNetwork.Rd
@@ -7,7 +7,8 @@
 plotIntegratedPatientNetwork(
   dataList,
   groupList,
-  makeNetFunc,
+  makeNetFunc = NULL,
+  sims = NULL,
   setName = "predictor",
   prune_pctX = 0.05,
   prune_useTop = TRUE,
@@ -33,6 +34,8 @@ clinical) and inner list to features to generate from that datatype.}
 
 \item{makeNetFunc}{(function) function to create features}
 
+\item{sims}{(list) rules for creating PSN. Preferred over makeNetFunc}
+
 \item{setName}{(char) name to assign the network in Cytoscape}
 
 \item{prune_pctX}{(numeric between 0 and 1) fraction of most/least 
diff --git a/man/predict.Rd b/man/predict.Rd
index 57747a8e..78b8966a 100644
--- a/man/predict.Rd
+++ b/man/predict.Rd
@@ -8,10 +8,10 @@ predict(
   trainMAE,
   testMAE,
   groupList,
-  featSel,
-  makeNetFunc,
+  selectedFeatures,
+  makeNetFunc = NULL,
+  sims = NULL,
   outDir,
-  impute = FALSE,
   verbose = FALSE,
   numCores = 1L,
   JavaMemory = 4L,
@@ -25,17 +25,17 @@ predict(
 
 \item{groupList}{(list) list of features used to train the model. Keys are data types, and values are lists for groupings within those datatypes.
 e.g. keys could include {'clinical','rna','methylation'}, and values within 'rna' could include pathway names {'cell cycle', 'DNA repair'}, etc.,
-featSel will be used to subset}
+selectedFeatures will be used to subset}
 
-\item{featSel}{(list) selected features to be used in the predictive model. 
+\item{selectedFeatures}{(list) selected features to be used in the predictive model. 
 keys are patient labels (e.g. "responder/nonresponder"), and values are feature names 
 identified by running buildPredictor(). Feature names must correspond to names of groupList, from which they will be subset.}
 
 \item{makeNetFunc}{(function) function to create PSN features from patient data. See makeNetFunc in buildPredictor() for details}
 
-\item{outDir}{(char) directory for results}
+\item{sims}{(list) rules for creating PSN. Preferred over makeNetFunc.}
 
-\item{impute}{(logical) if TRUE imputes train and test samples separately before creating features. Currently unsupported.}
+\item{outDir}{(char) directory for results}
 
 \item{verbose}{(logical) print messages}
 
diff --git a/man/psn__builtIn.Rd b/man/psn__builtIn.Rd
new file mode 100644
index 00000000..edd3ab82
--- /dev/null
+++ b/man/psn__builtIn.Rd
@@ -0,0 +1,21 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/similarities.R
+\name{psn__builtIn}
+\alias{psn__builtIn}
+\title{make PSN for built-in similarity functions}
+\usage{
+psn__builtIn(settings, verbose, ...)
+}
+\arguments{
+\item{settings}{(list) from makeNetFunc}
+
+\item{verbose}{(logical) print messages}
+
+\item{...}{parameters for makePSN_NamedMatrix()}
+}
+\value{
+(char) names of networks created. Side effect of network creation.
+}
+\description{
+make PSN for built-in similarity functions
+}
diff --git a/man/psn__corr.Rd b/man/psn__corr.Rd
new file mode 100644
index 00000000..c6d19968
--- /dev/null
+++ b/man/psn__corr.Rd
@@ -0,0 +1,21 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/similarities.R
+\name{psn__corr}
+\alias{psn__corr}
+\title{wrapper for PSNs using Pearson correlation}
+\usage{
+psn__corr(settings, verbose, ...)
+}
+\arguments{
+\item{settings}{(list) from makeNetFunc}
+
+\item{verbose}{(logical) print messages}
+
+\item{...}{parameters for makePSN_NamedMatrix()}
+}
+\value{
+(char) names of networks created. Side effect of network creation.
+}
+\description{
+wrapper for PSNs using Pearson correlation
+}
diff --git a/man/psn__custom.Rd b/man/psn__custom.Rd
new file mode 100644
index 00000000..8fb9a0be
--- /dev/null
+++ b/man/psn__custom.Rd
@@ -0,0 +1,23 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/similarities.R
+\name{psn__custom}
+\alias{psn__custom}
+\title{make PSN for custom similarity functions}
+\usage{
+psn__custom(settings, fn, verbose, ...)
+}
+\arguments{
+\item{settings}{(list) from makeNetFunc}
+
+\item{fn}{(function) custom similarity function}
+
+\item{verbose}{(logical) print messages}
+
+\item{...}{parameters for makePSN_NamedMatrix()}
+}
+\value{
+(char) names of networks created. Side effect of network creation.
+}
+\description{
+make PSN for custom similarity functions
+}
diff --git a/man/subsampleValidationData.Rd b/man/subsampleValidationData.Rd
new file mode 100644
index 00000000..588d3138
--- /dev/null
+++ b/man/subsampleValidationData.Rd
@@ -0,0 +1,31 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/splitTestTrain_partition.R
+\name{subsampleValidationData}
+\alias{subsampleValidationData}
+\title{Subsample a hold-out set from a larger patient dataset}
+\usage{
+subsampleValidationData(dataMAE, pctValidation = 0.2, verbose = TRUE)
+}
+\arguments{
+\item{dataMAE}{(MultiAssayExperiment) patient data to be subsampled. Must have columns ID (patient ID)
+and STATUS}
+
+\item{pctValidation}{(numeric) Fraction of dataset to include in the validation set. Value from 0.05 to 0.95.}
+
+\item{verbose}{(logical)}
+}
+\value{
+(list) Keys are trainMAE and validationMAE. These contain corresponding MultiAssayExperiments for
+training and test data
+}
+\description{
+Subsample a hold-out set from a larger patient dataset
+}
+\details{
+Creates a partition of data to be used for model validation after initial model building.
+In netDx, buildPredictor() is used for model training, and selected features from this exercise
+are used to validate a held-out dataset with the predict() function. Note that this function
+identifies a random subsample, which may result in a validation sample that is not representative
+of your training bias. If this method is used, please use data exploration techniques (e.g. UMAP)
+to ensure that validation accuracy is not confounded by stratification.
+}
diff --git a/tests/testthat/pathway_ex3.gmt b/tests/testthat/pathway_ex3.gmt
new file mode 100644
index 00000000..9073b6e1
--- /dev/null
+++ b/tests/testthat/pathway_ex3.gmt
@@ -0,0 +1,3 @@
+STEARATE BIOSYNTHESIS I (ANIMALS)%HUMANCYC%PWY-5972	stearate biosynthesis I (animals)	ELOVL1	ACOT7	ACSL1	ACSL5	ELOVL6	ACSL4	ACSL3	ACOT2	ACOT1	ACSBG1	ACSBG2	SLC27A2	ACOT4	
+PUTRESCINE DEGRADATION III%HUMANCYC%PWY-0	putrescine degradation III	ALDH3A2	ALDH3B2	ALDH3A1	ALDH1B1	MAOB	ALDH2	MAOA	ALDH3B1	SAT2	SAT1	
+TRYPTOPHAN DEGRADATION III (EUKARYOTIC)%HUMANCYC%TRYPTOPHAN-DEGRADATION-1	tryptophan degradation III (eukaryotic)	ACAT1	HADHB	GCDH	TDO2	KYNU	HAAO	AFMID	KMO	ACAA1	ACAT2	ACMSD	
diff --git a/tests/testthat/prepare_data.R b/tests/testthat/prepare_data.R
new file mode 100755
index 00000000..761dc1b7
--- /dev/null
+++ b/tests/testthat/prepare_data.R
@@ -0,0 +1,48 @@
+# setup brca data
+prepareData <- function(dat, setBinary=FALSE) {
+### clean up stage variable
+staget <- sub("[abcd]","",sub("t","",colData(dat)$pathology_T_stage))
+staget <- suppressWarnings(as.integer(staget))
+colData(dat)$STAGE <- staget
+
+### remove NA PAM50 calls, remove normal samples
+tmp <- colData(dat)$PAM50.mRNA
+if (!setBinary){
+	idx <- which(tmp %in% c("Normal-like","HER2-enriched"))
+} else {
+	idx <- union(which(tmp %in% c("Normal-like","HER2-enriched","Luminal B")),
+			which(is.na(staget)))
+}
+idx <- union(idx, which(is.na(tmp)))
+pID <- colData(dat)$patientID
+tokeep <- setdiff(pID, pID[idx])
+dat <- dat[,tokeep,]
+pam50 <- colData(dat)$PAM50.mRNA
+
+### where a patient has multiple instances of the same assay
+### just keep the first instance encountered
+smp <- sampleMap(dat)
+expr <- assays(dat)
+for (k in 1:length(expr)) {
+	samps <- smp[which(smp$assay==names(expr)[k]),]
+	notdup <- samps[which(!duplicated(samps$primary)),"colname"]
+	#message(sprintf("%s: %i notdup", names(expr)[k], length(notdup)))
+	dat[[k]] <- suppressMessages(dat[[k]][,notdup])
+}
+
+### create ID, STATUS columns, remove spaces/hyphens from patient labels
+pID <- colData(dat)$patientID
+colData(dat)$ID <- pID
+colData(dat)$STATUS <- pam50
+colData(dat)$STATUS <- gsub(" ",".",colData(dat)$STATUS)
+colData(dat)$STATUS <- gsub("-",".",colData(dat)$STATUS)
+
+if (setBinary){
+	st <- colData(dat)$STATUS
+	st[which(!st %in% "Luminal.A")] <- "other"
+	colData(dat)$STATUS <- st
+}
+
+return(dat)
+}
+
diff --git a/tests/testthat/test_buildpredictor.R b/tests/testthat/test_buildpredictor.R
index 1678d8a2..d0c19c65 100644
--- a/tests/testthat/test_buildpredictor.R
+++ b/tests/testthat/test_buildpredictor.R
@@ -26,7 +26,7 @@ test_that("split test/train works", {
 
 })
 
-test_that("feature construction and compilation",{
+test_that("feature construction and compilation works",{
 	# 20 patients, 10 case, 10 control
 	pheno <- data.frame(ID=sprintf("PAT%i",1:20),
 		STATUS=rep(c("case","control"),each=10))
@@ -69,12 +69,13 @@ test_that("feature construction and compilation",{
 	# directory contains GENES.TXT, NETWORKS.TXT INTERACTIONS folder
 	 outDir <- tempdir()
 	netDir <- sprintf("%s/tmp",outDir)
+	if (file.exists(netDir)) unlink(netDir,recursive=TRUE)
 	dir.create(netDir)
 
 	pheno_id <- setupFeatureDB(pheno,netDir)
 	x <-createPSN_MultiData(dataList=dataList,groupList=groupList,
 			pheno=pheno_id,
-			netDir=netDir,customFunc=makeNets,numCores=1,
+			netDir=netDir,makeNetFunc=makeNets,numCores=1,
 			verbose=FALSE)
 	
 	# number of nets equals those submitted for creation
@@ -102,35 +103,53 @@ test_that("feature construction and compilation",{
 	expect_match(dir(dbDir$dbDir,"lucene.index"),"lucene.index") # copied ok
 })
 
-###test_that("making PSN works: Pearson", {
-###	# create network
-###  # read in file
-###  # take two patients, compute pearson, check with outptu in file
-###	# num .profile files are greater than zero
-###})
-
-###test_that("making PSN works: custom func", {
-###	# same
-###	# num _cont.txt files are greater than zero
-###})
-###
-#### ------------------------------------------
-#### feature selection
-#### ------------------------------------------
-###
-###test_that("queries are made correctly", {
-###	# patients are subset of incPat
-###})
-###
-###test_that("feature selection gives expected output", {
-###	# one dir per class
-###	# N PRANK and NRANK files
-###	# correct number of _pathwayCV_score.txt files
-###})
-###
-#### ------------------------------------------
-#### predicting test label
-#### ------------------------------------------
-###test_that("test patients are labelled correctly", {
-###	# are all test patients labelled?
-###})
+test_that("holding out and validation set prediction works as it should", {
+
+	data(xpr,pheno)
+	pheno$AGE <- runif(nrow(pheno),min=10,max=45)
+	pathList <- list(
+		a=sample(rownames(xpr),25,F),
+		b=sample(rownames(xpr),100,F),
+		c=sample(rownames(xpr),30,F)
+	)
+	suppressMessages(require(MultiAssayExperiment))
+	objlist <- list(a=SummarizedExperiment(xpr))
+	mae <- MultiAssayExperiment(objlist,pheno)
+	
+	# test validation set features
+	dset <- subsampleValidationData(mae,pctValidation=0.2)
+	testthat::expect_equal(nrow(colData(dset[[1]])) + nrow(colData(dset[[2]])),
+				nrow(colData(mae)) )
+	tst <- colData(dset[[2]])$STATUS
+	lbl <- unique(tst)
+	testthat::expect_equal(sum(tst == lbl[1]), sum(tst==lbl[2]))
+	testthat::expect_equal(0.2, round(length(tst)/nrow(colData(mae)),digits=1) )
+
+	outDir <- tempdir()
+	if (file.exists(outDir)) unlink(outDir,recursive=TRUE)
+	dir.create(outDir)
+	featSel <- list(LumA=c("a.profile","age_cont.txt"),notLumA=c("a.profile","b.profile"))
+	predModel <- suppressMessages(
+  		predict(trainMAE=dset[[1]], testMAE=dset[[2]], 
+    		groupList=list(a=pathList,clinical=list(age="AGE")), 
+    		selectedFeatures=featSel,
+    		sims=list(a="pearsonCorr",clinical="normDiff"),
+    		outDir=outDir, verbose = FALSE)
+	)
+
+	for (g in c("LumA","notLumA")){
+		luma <- read.delim(sprintf("%s/%s/%s_query",outDir,g,g),sep=":")
+		pat <- unlist(strsplit(trimws(luma[1,1]),"\t")) 
+		# check that query patients are training samples for the corresponding class
+		testthat::expect_equal(sort(pat),sort(colData(dset[[1]])$ID[which(colData(dset[[1]])$STATUS==g)]))
+		# check that only selected features are used
+		nets <- read.delim(sprintf("%s/%s/networks/NETWORKS.txt",outDir,g),sep="\t",h=F)
+		testthat::expect_identical(featSel[[g]],nets[,2])
+		# test that only and all holdout set patients are being classified
+		dat <- read.delim(sprintf("%s/%s/%s_query-results.report.txt.PRANK",outDir,g,g),h=T,skip=1,sep="\t")
+		sb <- dat[which(!is.na(dat[,2])),]
+		testthat::expect_equal(nrow(colData(dset[[2]])), length(intersect(sb[,1],colData(dset[[2]])$ID)))		
+	}
+
+})
+
diff --git a/tests/testthat/test_suite.R b/tests/testthat/test_misc.R
similarity index 99%
rename from tests/testthat/test_suite.R
rename to tests/testthat/test_misc.R
index 8719dac4..7ead8a77 100644
--- a/tests/testthat/test_suite.R
+++ b/tests/testthat/test_misc.R
@@ -10,7 +10,6 @@ test_that("readPathways works", {
 		expect_lt(max(ln),201)
 })
 
-
 ###test_that("lasso filtering works", {
 ###	# make own subroutine
 ###})
diff --git a/tests/testthat/test_psn_construction.R b/tests/testthat/test_psn_construction.R
new file mode 100644
index 00000000..9778e819
--- /dev/null
+++ b/tests/testthat/test_psn_construction.R
@@ -0,0 +1,126 @@
+
+
+###test_that("invalid sims is flagged",{
+###
+###    expect_equal(TRUE, checkSimValid(list(a="pearsonCorr")))
+###    expect_identical(TRUE, checkSimValid(list(a="pearsonCorr",b=function(x) 2+4)))
+###    expect_identical(TRUE, checkSimValid(list(a="normDiff")))
+###    expect_error(checkSimValid(list(a="normDifff")))
+###    expect_error(checkSimValid(list(a=2)))
+###})
+
+#test_that("psns constructed using provided sims",{
+    rm(list=ls())
+    library(curatedTCGAData)
+    library(netDx)
+    brca <- suppressMessages(
+    curatedTCGAData("BRCA",
+               c("mRNAArray","RPPA*","Methylation_methyl27*"),
+	dry.run=FALSE,version="1.1.38"))
+    
+    source("prepare_data.R")
+    brca <- prepareData(brca,setBinary=TRUE)
+
+    groupList <- list()
+
+    # genes in mRNA data are grouped by pathways
+    pathFile <- "pathway_ex3.gmt"
+    pathList <- readPathways(pathFile)
+    groupList[["BRCA_mRNAArray-20160128"]] <- pathList
+    # clinical data is not grouped; each variable is its own feature
+    groupList[["clinical"]] <- list(
+          age="patient.age_at_initial_pathologic_diagnosis",
+    	   stage="STAGE"
+    )
+    # for methylation generate one feature containing all probes
+    # same for proteomics data
+    for (k in 2:3){
+    tmp <- list(rownames(experiments(brca)[[k]]));
+    names(tmp) <- names(brca)[k]
+    groupList[[names(brca)[k]]] <- tmp
+    }
+
+dataList <- dataList2List(brca,groupList)
+pheno <- data.frame(INTERNAL_ID=1:nrow(colData(brca)),
+    ID=colData(brca)$ID)
+
+netDir <- paste(tempdir(),"nets",sep=getFileSep())
+# ----------
+message("test pearson and built-in function")
+if (file.exists(netDir)) unlink(netDir,recursive=TRUE)
+dir.create(netDir,recursive=TRUE)
+sims <- list(a="pearsonCorr",b="normDiff",
+        c="pearsonCorr",d="pearsonCorr")
+names(sims) <- names(groupList)
+x <- createNetFuncFromSimList(dataList$assays,groupList,
+    netDir,sims)
+prof <- dir(netDir,".profile$")
+cont <- dir(netDir,"_cont.txt")
+testthat::expect_equal(7,length(x))
+pathways <- c(names(groupList[[1]]),names(groupList[3:4]))
+testthat::expect_identical(sort(pathways), sort(sub(".profile$","",prof)))
+print(all.equal(sort(names(groupList[[2]])),sort(sub("_cont.txt","",cont))))
+
+# ----------
+message("just pearson")
+gp1 <- groupList[1]; d1 <- dataList$assays[1]
+sims <- list(a="pearsonCorr"); names(sims) <- names(gp1)
+if (file.exists(netDir)) unlink(netDir,recursive=TRUE)
+dir.create(netDir,recursive=TRUE)
+x <- createNetFuncFromSimList(d1,gp1,
+    netDir,sims)
+prof <- dir(netDir,".profile$")
+cont <- dir(netDir,"_cont.txt")
+testthat::expect_equal(length(prof),3)
+testthat::expect_equal(length(cont),0)
+
+message("just built-in")
+gp1 <- groupList[2]; d1 <- dataList$assays["clinical"]
+sims <- list(a="normDiff"); names(sims) <- names(gp1)
+if (file.exists(netDir)) unlink(netDir,recursive=TRUE)
+dir.create(netDir,recursive=TRUE)
+x <- createNetFuncFromSimList(d1,gp1,
+    netDir,sims)
+prof <- dir(netDir,".profile$")
+cont <- dir(netDir,"_cont.txt")
+testthat::expect_equal(length(prof),0)
+testthat::expect_equal(length(cont),2)
+
+message("just custom")
+gp1 <- groupList[2]; d1 <- dataList$assays["clinical"]
+sims <- list(a=normDiff); names(sims) <- names(gp1)
+if (file.exists(netDir)) unlink(netDir,recursive=TRUE)
+dir.create(netDir,recursive=TRUE)
+x <- createNetFuncFromSimList(d1,gp1,
+    netDir,sims)
+prof <- dir(netDir,".profile$")
+cont <- dir(netDir,"_cont.txt")
+testthat::expect_equal(length(prof),0)
+testthat::expect_equal(length(cont),2)
+
+message("test full net creation function")
+sims <- list(a="pearsonCorr",b="normDiff",
+        c="pearsonCorr",d="pearsonCorr")
+names(sims) <- names(groupList)
+if (file.exists(netDir)) unlink(netDir,recursive=TRUE)
+dir.create(netDir,recursive=TRUE)
+netList<- createPSN_MultiData(
+    dataList=dataList$assays,
+    groupList=groupList,
+    pheno=pheno, netDir=netDir,
+    makeNetFunc=NULL, sims=sims,
+    numCores=1,verbose=TRUE)
+expect_equal(7, length(netList))
+testthat::expect_equal(2, length(dir(sprintf("%s/INTERACTIONS",netDir))))
+
+
+# test full predictor
+#set.seed(42) # make results reproducible
+##out <- 
+##   buildPredictor(dataList=brca,groupList=groupList,
+##      sims=sims,
+##      outDir=outDir, ## netDx requires absolute path
+##      numSplits=2L, featScoreMax=2L, featSelCutoff=1L,
+##	  numCores=nco)
+##    
+#})
\ No newline at end of file
diff --git a/vignettes/BuildPredictor.Rmd b/vignettes/BuildPredictor.Rmd
deleted file mode 100755
index 7d0fa61c..00000000
--- a/vignettes/BuildPredictor.Rmd
+++ /dev/null
@@ -1,329 +0,0 @@
----
-title: "Building binary classifier from clinical and 'omic data using pathway-level features"
-author: "Shraddha Pai"
-package: netDx
-date: "`r Sys.Date()`"
-output: 
-  BiocStyle::html_document:
-    toc_float: true
-vignette: >
-    %\VignetteIndexEntry{01. Build binary predictor and view performance, top features and integrated Patient Similarity Network}.
-    %\VignetteEngine{knitr::rmarkdown}
-    %\VignetteEncoding{UTF-8}
----
-
-# Introduction
-
-In this example we will build a predictor to classify breast tumours as being either of Luminal A subtype or otherwise. The process is identical for classifying three or more labels, and the example uses minimal data for quick runtime.
-
-For this we will use data from the  The Cancer Genome Atlas, and will integrate three types of -omic data: 
-
-* gene expression from Agilent mRNA microarrays and
-* miRNA sequencing 
-
-
-```{r, fig.cap="We will integrate two layers of genomic data. Each layer will be converted into a single patient similarity network using Pearson correlation for pairwise similarity. ",echo=FALSE, crop=NULL}
-knitr::opts_chunk$set(crop=NULL)
-knitr::include_graphics("images/vignette1_design.jpg")
-```
-
-## Get and prepare data
-
-In this example, we use curated multi-modal data from The Cancer Genome Atlas, gotten from the BioConductor `curatedTCGAData` package. Data for all cancer types profiled in TCGA are available through this package; [see this tutorial for details](https://bioconductor.org/packages/release/data/experiment/vignettes/curatedTCGAData/inst/doc/curatedTCGAData.html). 
-
-```{r, eval=TRUE}
-suppressMessages(library(curatedTCGAData))
-```
-
-# Data 
-
-In this example, we use curated data from The Cancer Genome Atlas, through the BioConductor `curatedTCGAData` package. The goal is to classify a breast tumour into either a Luminal A subtype or otherwise. The predictor will integrate clinical variables selected by the user, along with gene expression data.
-
-Here we load the required packages and download clinical and gene expression data.
-```{r,eval=TRUE}
-suppressMessages(library(curatedTCGAData))
-```
-
-Let's take a look at the available data for breast cancer, without downloading any (set `dry.run=TRUE`). 
-Note that the new release of BioConductor (3.13) actually allows users to fetch one of two versions of TCGA data. 
-
-```{r, eval=TRUE}
-curatedTCGAData(diseaseCode="BRCA", assays="*",dry.run=TRUE, version="1.1.38")
-```
-Now let's actually download the data, getting just the three layers we need:
-
-```{r, eval=TRUE}
-brca <- suppressMessages(curatedTCGAData("BRCA",
-                                         c("mRNAArray", 
-										 "miRNASeqGene"),
-                                         dry.run=FALSE, version="1.1.38"))
-```
-
-This call returns a `MultiAssayExperiment` object. Recall that this is a container for storing multiple assays performed on the same set of samples. [See this tutorial](https://bioconductor.org/packages/release/bioc/vignettes/MultiAssayExperiment/inst/doc/QuickStartMultiAssay.html) to learn more.
-
-Let's briefly explore the `brca` `MultiAssayExperiment` object.
-
-```{r, eval=TRUE}
-brca
-```
-
-`assays()` returns a `list` with all -omic data associated with this object.  
-
-```{r, eval=TRUE}
-summary(assays(brca))
-```
-`names()` shows the datatypes in each slot of `assays()`:
-
-```{r, eval=TRUE}
-names(assays(brca))
-```
-
-So miRNA data is in slot 1, gene expression in slot 2, etc., 
-
-We can subset the data to see what it looks like. Let's do that for the miRNA data, looking at just the first five measures:
-
-```{r, eval=TRUE}
-mir <- assays(brca)[["BRCA_miRNASeqGene-20160128"]]
-head(mir[,1:5])
-```
-
-Patient metadata is contained in the `colData()` slot. Rows contain data for each patient and columns contain measures such as clinical characteristics:
-
-```{r, eval=TRUE}
-pheno <- colData(brca)
-colnames(pheno)[1:20]
-head(pheno[,1:5])
-```
-
-This next code block prepares the TCGA data. This includes:
-
-* removing duplicate samples
-* reformatting patient IDs (e.g. removing spaces and hyphens)
-* creating an `ID` column in `colData(brca)`, which contains unique patient IDs
-* creating a `STATUS` column in `colData(brca)` which contains the patient labels (i.e what we want netDx to classify). 
-
-
-In practice you would prepare the dataset once and save it to file, then separately load it before running netDx; i.e. decouple data processing and running the predictor. The data processing code has been moved into a supporting file, `prepare_data.R`. You can explore it after the lab to see how some things are achieved (e.g. removing duplicate samples). For now, let's just run it.
-
-```{r, eval=TRUE}
-source("prepare_data.R")
-brca <- prepareData(brca,setBinary=TRUE)
-```
-
-The important thing is to create `ID` and `STATUS` columns in the sample metadata table. netDx uses these to get the patient identifiers and labels, respectively.
-
-
-```{r, eval=TRUE}
-pheno <- colData(brca)
-head(pheno[,c("ID","STATUS")])
-table(pheno$STATUS,useNA="always")  # good practice: useNA="always" shows missing values
-```
-
-## Create feature design rules
-
-Now let's set up the data for input to netDx. 
-
-netDx allows the user to define how data is converted into patient similarity networks (or PSNs), which are the features that go into the model. This is done specifically by telling the model how to:
-
-* **group** different types of data and 
-* **define similarity** for each of these (e.g. Pearson correlation, normalized difference, etc.,).
-
-The relevant input parameters are:
-
-* `groupList`: sets of input data that would correspond to individual networks (e.g. genes grouped into pathways)
-* `makeNets()`: an R function telling netDx what similarity metric to use for each data layer
-
-Let's start by loading the `netDx` package.
-
-
-```{r, eval=TRUE}
-suppressWarnings(suppressMessages(require(netDx)))
-```
-
-Let's set up each of the input arguments one by one.
-
-### groupList
-**What is this:** `groupList` tells netDx how to group measures within a layer, to generate a PSN. Measures could be individual genes, proteins, CpG bases (in DNA methylation data), clinical variables, etc., 
-
-In this simple example we just create a single PSN for each datatype, containing all measures from that datatype.
-
-
-```{r, eval=TRUE}
-expr <- assays(brca)
-groupList <- list()
-for (k in 1:length(expr)) {	# loop over all layers
-	cur <- expr[[k]]; nm <- names(expr)[k]
-
-	# all measures from this layer go into our single PSN
-	groupList[[nm]] <- list(nm=rownames(cur)) 
-
-	# assign same layer name as in input data
-	names(groupList[[nm]])[1] <- nm;
-}
-```
-Notice that `groupList` is a two tiered list, or list-of-lists. The first tier is for each data layers, with names matching those in `assays(brca)`. The second tier contains all the PSNs we want to make for that layer. In this lab exercise we create only one PSN per data layer, using all the measures from an -omic assay. e.g. One PSN based on similarity across entire transcriptome, one for methylome, etc.,). So in this lab exercise, the inner tier simply contains one entry, with all measures for the given layer. 
-
-This design will get more interesting in Lab 2, when we create pathway-level features. 
-
-Let's take a look at `groupList`. Here is the first tier:
-
-
-```{r, eval=TRUE}
-summary(groupList)
-```
-
-And the second tier:
-
-```{r, eval=TRUE}
-names(groupList[["BRCA_mRNAArray-20160128"]])
-length(groupList[["BRCA_mRNAArray-20160128"]][[1]])
-head(groupList[["BRCA_mRNAArray-20160128"]][[1]])
-```
-
-### Define patient similarity for each network
-The `makeNets` function tells the predictor how to create networks from provided input data.
-
-This function requires `dataList`,`groupList`, and `netDir` as input variables. The residual `...` parameter is to pass additional variables to `makePSN_NamedMatrix()`, notably `numCores` (number of parallel jobs).
-
-netDx requires that this function have:
-
-* `dataList`,`groupList`, and `netDir` as input variables. The residual `...` parameter is to pass additional variables to `makePSN_NamedMatrix()`, notably number of cores for parallel processing (`numCores`). 
-
-
-```{r,  eval=TRUE}
-makeNets <- function(dataList, groupList, netDir,...) {
-	netList <- c() # initialize before is.null() check
-	
-	layerNames <- c("BRCA_miRNASeqGene-20160128",
-		"BRCA_mRNAArray-20160128")
-	
-	for (nm in layerNames){  			## for each layer
-		if (!is.null(groupList[[nm]])){ ## must check for null for each layer
-			netList_cur <- makePSN_NamedMatrix(
-				dataList[[nm]],
-				rownames(dataList[[nm]]),	## names of measures (e.g. genes, CpGs)
-				groupList[[nm]],			## how to group measures in that layer
-				netDir,						## leave this as-is, netDx will figure out where this is.
-				verbose=FALSE, 			
-				writeProfiles=TRUE,   		## use Pearson correlation-based similarity
-				...
-				)
-
-			netList <- c(netList,netList_cur)	## just leave this in
-		}
-	}
-	return(unlist(netList))	## just leave this in 
-}
-```
-
-## Build predictor
-
-Finally we call the function that runs the netDx predictor. We provide:
-
-* patient data  (`dataList`)
-* grouping rules (`groupList`)
-* function to create PSN from data, includes choice of similarity metric (`makeNetFunc`)
-* number of train/test splits over which to collect feature scores and average performance (`numSplits`), 
-* maximum score for features in one round of feature selection  (`featScoreMax`, set to 10)
-* threshold to call feature-selected networks for each train/test split (`featSelCutoff`); only features scoring this value or higher will be used to classify test patients,
-* number of cores to use for parallel processing (`numCores`).
-
-The call below runs 10 train/test splits. 
-Within each split, it:
-
-* splits data into train/test using the default split of 80:20 (`trainProp=0.8`)
-* score networks between 0 to 10 (i.e. `featScoreMax=10L`)
-* uses networks that score >=9 out of 10 (`featSelCutoff=9L`) to classify test samples for that split.
-
-In practice a good starting point is `featScoreMax=10`, `featSelCutoff=9` and `numSplits=10L`, but these parameters depend on the sample sizes in the dataset and heterogeneity of the samples. 
-
-This step can take a few hours based on the current parameters, so we're commenting this out for the tutorial and will simply load the results.
- 
-```{r lab1-buildpredictor ,eval=TRUE}
-nco <- round(parallel::detectCores()*0.75) # use 75% available cores
-message(sprintf("Using %i of %i cores", nco, parallel::detectCores()))
-t0 <- Sys.time()
-set.seed(42) # make results reproducible
-outDir <- paste(tempdir(),randAlphanumString(),
-	"pred_output",sep=getFileSep())
-if (file.exists(outDir)) unlink(outDir,recursive=TRUE)
-model <- suppressMessages(buildPredictor(
-	dataList=brca,			## your data
-	groupList=groupList,	## grouping strategy
-	makeNetFunc=makeNets,	## function to build PSNs
-	outDir=outDir, 			## output directory
-	trainProp=0.8,			## pct of samples to use to train model in
-							## each split
-	numSplits=2L,			## number of train/test splits
- featSelCutoff=1L,		## threshold for calling something
-							## feature-selected
-	featScoreMax=2L,		## max score for feature selection
- numCores=nco,			## set higher for parallelizing
- debugMode=FALSE,
- keepAllData=FALSE,	## set to TRUE for debugging or low-level files used by the dictor
- logging="none"
-  ))
-t1 <- Sys.time()
-print(t1-t0)
-```
-
-## Examine results
-
-Now we get model output, including performance for various train/test splits and consistently high-scoring features. 
-
-
-In the function below, we define top-scoring features as those which score two out of two in at least half of the train/test splits:
-
-```{r lab1-getresults,eval=TRUE}
-results <- getResults(model,unique(colData(brca)$STATUS),
-	featureSelCutoff=2L,featureSelPct=0.50)
-```
-
-`results` contains `performance`, `selectedFeatures` for each patient label, and the table of feature `scores`.
-
-```{r, eval=TRUE}
-summary(results)
-```
-
-Look at the performance:
-```{r, eval=TRUE}
-results$performance
-```
-
-Look at feature scores for all labels, across all train-test splits:
-
-```{r,  eval=TRUE}
-results$featureScores
-```
-
-Let's examine our confusion matrix:
-```{r, eval=TRUE}
-confMat <- confusionMatrix(model)
-```
-
-*Note: Rows of this matrix don't add up to 100% because the matrix is an average of the confusion matrices from all of the train/test splits.*
-
-And here are selected features, which are those scoring 2 out of 2 in at least half of the splits. This threshold is simply for illustration. In practice we would run at least 10 train/test splits (ideally 100+), and look for features that score 7+ out of 10 in >70% splits.
-
-```{r, eval=TRUE}
-results$selectedFeatures
-```
-
-We finally get the integrated PSN and visualize it using a tSNE plot:
-
-```{r, fig.width=8,fig.height=8, eval=TRUE}
-## this call doesn't work in Rstudio; for now we've commented this out and saved the PSN file. 
-psn <- getPSN(brca,groupList,makeNets,results$selectedFeatures)
-
-require(Rtsne)
-tsne <- tSNEPlotter(
-	psn$patientSimNetwork_unpruned, 
-	colData(brca)
-	)
-```
-
-
-## sessionInfo
-```{r}
-sessionInfo()
-```
diff --git a/vignettes/ThreeWayClassifier.Rmd b/vignettes/ThreeWayClassifier.Rmd
index 4b532ba7..e817e7b8 100755
--- a/vignettes/ThreeWayClassifier.Rmd
+++ b/vignettes/ThreeWayClassifier.Rmd
@@ -64,6 +64,14 @@ pID <- colData(brca)$patientID
 colData(brca)$ID <- pID
 ```
 
+ADD HOLDOUT TEXT HERE
+```{r}
+set.seed(123)
+dsets <- subsampleValidationData(brca,pctValidation=0.1)
+brca <- dsets$trainMAE
+holdout <- dsets$validationMAE
+```
+
 # Rules to create features (patient similarity networks)
 
 Our plan is to group gene expression data by pathways and clinical data by single variables. We will treat methylation and proteomic data each as a single feature, so each of those groups will contain the entire input table for those corresponding data types. 
@@ -97,49 +105,13 @@ groupList[[names(brca)[3]]] <- tmp
 
 ## Define patient similarity for each network
 
-We provide `netDx` with a custom function to generate similarity networks (i.e. features). The first block tells netDx to generate correlation-based networks using everything but the clinical data. This is achieved by the call:
-```{r,eval=FALSE}
-makePSN_NamedMatrix(..., writeProfiles=TRUE,...)`
-```
-
-The second block makes a different call to `makePSN_NamedMatrix()` but this time, requesting the use of the normalized difference similarity metric. This is achieved by calling:
-```{r,eval=FALSE}
-   makePSN_NamedMatrix(,..., 
-                       simMetric="custom", customFunc=normDiff,
-                       writeProfiles=FALSE)
-```
+ADD SOME INTRO TEXT HERE
 
 `normDiff` is a function provided in the `netDx` package, but the user may define custom similarity functions in this block of code and pass those to `makePSN_NamedMatrix()`, using the `customFunc` parameter.
 
 ```{r,eval=TRUE}
-makeNets <- function(dataList, groupList, netDir,...) {
-	netList <- c() # initialize before is.null() check
-	# correlation-based similarity for mRNA, RPPA and methylation data
-	# (Pearson correlation)
-	for (nm in setdiff(names(groupList),"clinical")) {
-	   # NOTE: the check for is.null() is important!
-		if (!is.null(groupList[[nm]])) {
-		netList <- makePSN_NamedMatrix(dataList[[nm]],
-		             rownames(dataList[[nm]]),
-                   groupList[[nm]],netDir,verbose=FALSE,
-		             writeProfiles=TRUE,...) 
-		}
-	}
-	
-	# make clinical nets (normalized difference)
-	netList2 <- c()
-	if (!is.null(groupList[["clinical"]])) {
-	netList2 <- makePSN_NamedMatrix(dataList$clinical, 
-		rownames(dataList$clinical),
-		groupList[["clinical"]],netDir,
-		simMetric="custom",customFunc=normDiff, # custom function
-		writeProfiles=FALSE,
-		sparsify=TRUE,verbose=TRUE,...)
-	}
-	netList <- c(unlist(netList),unlist(netList2))
-	return(netList)
-}
-
+sims <- list(a="pearsonCorr",b="normDiff",c="pearsonCorr",d="pearsonCorr")
+names(sims) <- names(groupList)
 ```
 
 # Build predictor
@@ -147,64 +119,87 @@ makeNets <- function(dataList, groupList, netDir,...) {
 Finally we make the call to build the predictor. 
  
 ```{r,eval=TRUE}
-set.seed(42) # make results reproducible
-
-# location for intermediate work
-# set keepAllData to TRUE to not delete at the end of the 
-# predictor run.
-# This can be useful for debugging.
 nco <- round(parallel::detectCores()*0.75) # use 75% available cores
 message(sprintf("Using %i of %i cores", nco, parallel::detectCores()))
+
 outDir <- paste(tempdir(),"pred_output",sep=getFileSep()) # use absolute path
+if (file.exists(outDir)) unlink(outDir,recursive=TRUE)
 numSplits <- 2L
-out <- suppressMessages(
-   buildPredictor(dataList=brca,groupList=groupList,
-      makeNetFunc=makeNets,
-      outDir=outDir, ## netDx requires absolute path
-      numSplits=numSplits, featScoreMax=2L, featSelCutoff=1L,
-	  numCores=nco)
-)
+t0 <- Sys.time()
+model <- suppressMessages(
+buildPredictor(
+	dataList=brca,			## your data
+	groupList=groupList,	## grouping strategy
+	sims=sims,
+	outDir=outDir, 			## output directory
+	trainProp=0.8,			## pct of samples to use to train model in
+							## each split
+	numSplits=2L,			## number of train/test splits
+ 	featSelCutoff=1L,		## threshold for calling something
+							## feature-selected
+	featScoreMax=2L,		## max score for feature selection
+ numCores=nco,			## set higher for parallelizing
+ debugMode=FALSE,
+ keepAllData=FALSE,	## set to TRUE for debugging or low-level files used by the dictor
+ logging="none"
+  ))
+t1 <- Sys.time()
+print(t1-t0)
 ```
 
-# Examine output
-The results are stored in the list object returned by the `buildPredictor()` call.
-This list contains:
 
-* `inputNets`: all input networks that the model started with. 
-* `Split<i>`: a list with results for each train-test split
-  * `featureScores`: feature scores for each label (list with `g` entries, where `g` is number of patient labels). Each entry contains the feature selection scores for the corresponding label.
-  * `featureSelected`: vector of features that pass feature selection. List of length `g`, with one entry per label.
-  * `predictions`: real and predicted labels for test patients
-  * `accuracy`: percent accuracy of predictions
+```{r,eval=TRUE}
+results <- getResults(model,unique(colData(brca)$STATUS),
+	featureSelCutoff=2L,featureSelPct=0.50)
+```
+
+# Validate on independent samples
+Now we use `predict()` to classify samples in the independent dataset. We provide the model with feature design rules in `groupList`, the list of selected features to use in `featSelNet`, the function to convert data into patient similarity networks in `makeNets`, as well as the original and validated datasets in `brca` and `holdout` respectively.
+
+The training data needs to be provided because netDx creates a single patient similarity network with both training and test data. It then uses label propagation to "diffuse" patient labels from training samples to test samples, and labels the latter based on which class they are most similar to.
 
 ```{r,eval=TRUE}
-summary(out)
-summary(out$Split1)
+outDir <- paste(tempdir(), randAlphanumString(), 
+  sep = getFileSep())
+if (file.exists(outDir)) unlink(outDir,recursive=TRUE)
+dir.create(outDir)
+
+predModel <- suppressMessages(
+  predict(trainMAE=brca, testMAE=holdout, 
+    groupList=groupList, 
+    selectedFeatures=results$selectedFeatures, 
+    sims=sims,
+    outDir=outDir, verbose = FALSE)
+)
 ```
 
-Compute accuracy for three-way classificationL
+# Plot results of validation
+
+Finally we examine how well our model performed, using `getPerformance()`. 
 
+Compute performance:
 ```{r,eval=TRUE}
-# Average accuracy
-st <- unique(colData(brca)$STATUS) 
-acc <- matrix(NA,ncol=length(st),nrow=numSplits) 
-colnames(acc) <- st 
-for (k in 1:numSplits) { 
-	pred <- out[[sprintf("Split%i",k)]][["predictions"]];
-	tmp <- pred[,c("ID","STATUS","TT_STATUS","PRED_CLASS",
-	                 sprintf("%s_SCORE",st))]
-	for (m in 1:length(st)) {
-	   tmp2 <- subset(tmp, STATUS==st[m])
-	   acc[k,m] <- sum(tmp2$PRED==tmp2$STATUS)/nrow(tmp2)
-	}
-}
-print(round(acc*100,2))
+perf <- getPerformance(predModel, 
+  unique(colData(brca)$STATUS))
+
+message(sprintf("AUROC=%1.2f*100", perf$auroc))
+message(sprintf("AUPR=%1.2f*100", perf$aupr))
+message(sprintf("Accuracy=%1.1f%%", perf$acc))
 ```
-Also, examine the confusion matrix. We can see that the model perfectly classifies basal tumours, but performs poorly in distinguishing between the two types of luminal tumours. 
 
-```{r, eval=TRUE}
-res <- out$Split1$predictions
-print(table(res[,c("STATUS","PRED_CLASS")]))
+We plot the AUROC and AUPR curves using `plotPerf_multi()`.
+
+```{r,eval=TRUE}
+plotPerf_multi(list(perf$rocCurve),
+  plotTitle = sprintf(
+    "BRCA Validation: %i samples", 
+    nrow(colData(holdout))))
+plotPerf_multi(list(perf$prCurve), 
+  plotType = "PR",
+  plotTitle = sprintf(
+    "BRCA Validation: %i samples", 
+    nrow(colData(holdout))))
+
 ```
 
 # sessionInfo
diff --git a/vignettes/ValidateNew.Rmd b/vignettes/ValidateNew.Rmd.old
similarity index 97%
rename from vignettes/ValidateNew.Rmd
rename to vignettes/ValidateNew.Rmd.old
index 9ada52fb..1ce8d012 100755
--- a/vignettes/ValidateNew.Rmd
+++ b/vignettes/ValidateNew.Rmd.old
@@ -176,9 +176,11 @@ if (file.exists(outDir)) unlink(outDir,recursive=TRUE)
 dir.create(outDir)
 
 predModel <- suppressMessages(
-  predict(brca, holdout, groupList, 
-    results$selectedFeatures, makeNets,
-    outDir, verbose = FALSE)
+  predict(trainMAE=brca, testMAE=holdout, 
+    groupList=groupList, 
+    featSel=results$selectedFeatures, 
+    makeNetFunc=makeNets,
+    outDir=outDir, verbose = FALSE)
 )
 ```
 
diff --git a/vignettes/makeSimFunction.R b/vignettes/makeSimFunction.R
new file mode 100644
index 00000000..21d9e57b
--- /dev/null
+++ b/vignettes/makeSimFunction.R
@@ -0,0 +1,117 @@
+#' returns valid built-in similarity functions
+allowedSims <- function(){
+  c("pearson_corr","normDiff","avgNormDiff",
+        "sim.pearscale","sim.eucscale")
+}
+
+#' checks if provided similarity functions are valid. Returns error if not
+#'
+#' @param sims (list) keys are layer names, values are functions or characters 
+#' (names of built-in similarity functions)
+#' @return TRUE if all pass check. Else throws error.
+checkSimValid <- function(sims){
+    allowed <- allowedSims
+    for (k in names(sims)){
+        if (class(sims[[k]])!="function"){
+            if class(sims[[k]])!="character"{
+                stop(paste("Invalid sims datatype. ",
+                    "sims entries must be functions or keywords (characters) ",
+                    "for built-in similarity functions.",sep=""))
+            } else {
+                if (!sims[[k]] %in% allowed){
+                    stop(paste(
+                            sprintf("sims[[%s]] has invalid similarity type:",k),
+                            sims[[k]],". ",
+                            "Allowed values are: {%s}",
+                            paste(allowed,collapse=",")))
+                }
+            }
+        }
+    }
+    return(TRUE)
+}
+
+makeNetFunc <- function(dataList, groupList, netDir, sims,...){    
+    settings <- list(dataList=dataList,groupList=groupList,
+                    netDir=netDir,sims=sims)
+    netList <- c()    
+    for (nm in names(sims)){
+        csim <- sims[[nm]]
+        netList_cur <- NULL
+
+        cur_set <- settings; 
+        cur_set[["name"]] <- nm; cur_set[["similarity"]] <- csim
+
+        if (!is.null(groupList[[nm]])){
+            if (class(csim)=="function") {# custom function
+                netList_cur <- builtInPSN(cur_set,csim,...)
+            } else if (csim == "pearson_corr") {
+                netList_cur <- corrPSN(cur_set,...)
+            } else {
+                netList_cur <- builtInPSN(cur_set,...)
+            }
+            netList <- c(netList,netList_cur)
+        }
+    }
+    unlist(netList)
+}
+
+#' make PSN for built-in similarity functions
+#' @param settings (list) from makeNetFunc
+builtInPSN <- function(settings,...){
+funcs <- list(
+    "normDiff"=normDiff,
+    "avgNormDiff"=avgNormDiff,
+    "sim.pearscale"=sim.pearscale,
+    "sim.eucscale"=sim.eucscale
+)
+
+    message(sprintf("Layer %s: Function %s",settings$name,settings$similarity))
+
+    nm <- settings$name
+    netList <- makePSN_NamedMatrix(
+        settings$dataList[[nm]],
+		rownames(settings$dataList[[nm]]),
+		settings$groupList[[nm]],
+        settings$netDir,
+		simMetric="custom",
+        customFunc=funcs[[settings$similarity]], # custom function
+		writeProfiles=FALSE,
+		sparsify=TRUE,...
+    )
+    netList
+}
+
+#' make PSN for custom similarity functions
+#' @param settings (list) from makeNetFunc
+customPSN <- function(settings,fn, ...){
+    nm <- settings$name
+    message(sprintf("Layer %s: CUSTOM FUNCTION",settings$name))
+    netList <- makePSN_NamedMatrix(
+        settings$dataList[[nm]],
+		rownames(settings$dataList[[nm]]),
+		settings$groupList[[nm]],
+        settings$netDir,
+		simMetric="custom",customFunc=fn, # custom function
+		writeProfiles=FALSE,
+		sparsify=TRUE,...
+    )
+    netList
+}
+
+#' wrapper for PSNs using Pearson correlation
+#' @param settings (list) from makeNetFunc
+corrPSN <- function(settings,...){
+    message(sprintf("Layer %s: PEARSON CORR",settings$name))
+    nm <- settings$name
+    netList <- makePSN_NamedMatrix(
+				settings$dataList,
+				rownames(settings$dataList[[nm]]),	## names of measures (e.g. genes, CpGs)
+				settings$groupList[[nm]],			## how to group measures in that layer
+				settings$netDir,						## leave this as-is, netDx will figure out where this is.
+				verbose=FALSE, 			
+				writeProfiles=TRUE,   		## use Pearson correlation-based similarity
+				...
+				)
+    return(netList)
+}
\ No newline at end of file