@@ -4,6 +4,10 @@

   ## DelayedMatrixStats::rowSds() works for both base and 

   ## DelayedArray matrices

   sdGenes <- DelayedMatrixStats::rowSds(expr)

+  ## the following fixes this bug, see issues

+  ## https://github.com/rcastelo/GSVA/issues/54

+  ## https://github.com/HenrikBengtsson/matrixStats/issues/204

+  sdGenes[sdGenes < 1e-10] <- 0

   if (any(sdGenes == 0) || any(is.na(sdGenes))) {

     warning(sum(sdGenes == 0 | is.na(sdGenes)),

             " genes with constant expression values throuhgout the samples.")

@@ -66,27 +66,3 @@

   m@x <- unlist(x, use.names=FALSE)

   m

 }

-

-## filter out genes which non-zero values have

-## constant expression values

-.filterFeaturesSparse <- function(expr, method) {

-  

-  sdGenes <- sapply(.sparseToList(expr, 1), sd)

-  

-  if (any(sdGenes == 0) || any(is.na(sdGenes))) {

-    warning(sum(sdGenes == 0 | is.na(sdGenes)),

-            " genes with constant expression values throuhgout the samples.")

-    if (method != "ssgsea") {

-      warning("Since argument method!=\"ssgsea\", genes with constant expression values are discarded.")

-      expr <- expr[sdGenes > 0 & !is.na(sdGenes), ]

-    }

-  }

-  

-  if (nrow(expr) < 2)

-    stop("Less than two genes in the input assay object\n")

-  

-  if(is.null(rownames(expr)))

-    stop("The input assay object doesn't have rownames\n")

-  

-  expr

-}

@@ -28,20 +28,12 @@

 ## features in both input objects follow the same nomenclature,

 .mapGeneSetsToFeatures <- function(gsets, features) {

-  ## fastmatch::fmatch() modifies the 'table' argument (i.e., the

-  ## second argument) in place by adding the attribute '.match.hash'

-  ## https://github.com/rcastelo/GSVA/issues/39

-  ## to avoid that undesired feature we duplicate 'features'

-  ## by adding an "impossible value" at the end and let

-  ## fastmatch::match() work with the duplicated object

-  features2 <- c(features, "&!%impossiblevalue%!&")

+  ## Aaron Lun's suggestion at

+  ## https://github.com/rcastelo/GSVA/issues/39#issuecomment-765549620

+  gsets2 <- CharacterList(gsets)

+  mt <- match(gsets2, features)

+  mapdgenesets <- as.list(mt[!is.na(mt)])

-  ## map to the actual features for which expression data is available

-  mapdgenesets <- lapply(gsets,

-                         function(x, y)

-                           as.vector(na.omit(fastmatch::fmatch(x, y))),

-                         features2)

-  

   if (length(unlist(mapdgenesets, use.names=FALSE)) == 0)

     stop("No identifiers in the gene sets could be matched to the identifiers in the expression data.")

@@ -13,9 +13,41 @@

     }

   } 

+  if (nrow(expr) < 2)

+    stop("Less than two genes in the input assay object\n")

+  

+  if(is.null(rownames(expr)))

+    stop("The input assay object doesn't have rownames\n")

+  

   expr

 }

+## maps gene sets content in 'gsets' to 'features', where 'gsets'

+## is a 'list' object with character string vectors as elements,

+## and 'features' is a character string vector object. it assumes

+## features in both input objects follow the same nomenclature,

+.mapGeneSetsToFeatures <- function(gsets, features) {

+

+  ## fastmatch::fmatch() modifies the 'table' argument (i.e., the

+  ## second argument) in place by adding the attribute '.match.hash'

+  ## https://github.com/rcastelo/GSVA/issues/39

+  ## to avoid that undesired feature we duplicate 'features'

+  ## by adding an "impossible value" at the end and let

+  ## fastmatch::match() work with the duplicated object

+  features2 <- c(features, "&!%impossiblevalue%!&")

+

+  ## map to the actual features for which expression data is available

+  mapdgenesets <- lapply(gsets,

+                         function(x, y)

+                           as.vector(na.omit(fastmatch::fmatch(x, y))),

+                         features2)

+  

+  if (length(unlist(mapdgenesets, use.names=FALSE)) == 0)

+    stop("No identifiers in the gene sets could be matched to the identifiers in the expression data.")

+

+  mapdgenesets

+}

+

 ## transforms a dgCMatrix into a list of its

 ## non-zero values by MARGIN (1 for row, 2 for column)

 .sparseToList <-function(dgCMat, MARGIN){

@@ -58,5 +90,11 @@

     }

   }

+  if (nrow(expr) < 2)

+    stop("Less than two genes in the input assay object\n")

+  

+  if(is.null(rownames(expr)))

+    stop("The input assay object doesn't have rownames\n")

+  

   expr

-}

\ No newline at end of file

+}

@@ -1,6 +1,8 @@

 .filterFeatures <- function(expr, method) {

   ## filter out genes with constant expression values

+  ## DelayedMatrixStats::rowSds() works for both base and 

+  ## DelayedArray matrices

   sdGenes <- DelayedMatrixStats::rowSds(expr)

   if (any(sdGenes == 0) || any(is.na(sdGenes))) {

     warning(sum(sdGenes == 0 | is.na(sdGenes)),

@@ -1,7 +1,7 @@

 .filterFeatures <- function(expr, method) {

   ## filter out genes with constant expression values

-  sdGenes <- apply(expr, 1, sd)

+  sdGenes <- DelayedMatrixStats::rowSds(expr)

   if (any(sdGenes == 0) || any(is.na(sdGenes))) {

     warning(sum(sdGenes == 0 | is.na(sdGenes)),

             " genes with constant expression values throuhgout the samples.")

@@ -35,6 +35,12 @@

   result

 }

+.dgCapply<-function(m,f, MARGIN){

+  x <- lapply(.sparseToList(m, MARGIN), f)

+  m@x <- unlist(x, use.names=FALSE)

+  m

+}

+

 ## filter out genes which non-zero values have

 ## constant expression values

 .filterFeaturesSparse <- function(expr, method) {

@@ -13,3 +13,42 @@

   expr

 }

+

+## transforms a dgCMatrix into a list of its

+## non-zero values by MARGIN (1 for row, 2 for column)

+.sparseToList <-function(dgCMat, MARGIN){

+  MARGIN <- as.integer(MARGIN)

+  J <- rep(1:ncol(dgCMat), diff(dgCMat@p))

+  I <- dgCMat@i + 1

+  x <- dgCMat@x

+  if (MARGIN == 1L) {

+    result <- split(x, I)

+    names(result) <- rownames(dgCMat)[as.numeric(names(result))]

+  } else if (MARGIN == 2L) {

+    result <- split(x, J)

+    names(result) <- colnames(dgCMat)[as.numeric(names(result))]

+  }

+  else {

+    warning("invalid MARGIN; return NULL")

+    result <- NULL

+  }

+  result

+}

+

+## filter out genes which non-zero values have

+## constant expression values

+.filterFeaturesSparse <- function(expr, method) {

+  

+  sdGenes <- sapply(.sparseToList(expr, 1), sd)

+  

+  if (any(sdGenes == 0) || any(is.na(sdGenes))) {

+    warning(sum(sdGenes == 0 | is.na(sdGenes)),

+            " genes with constant expression values throuhgout the samples.")

+    if (method != "ssgsea") {

+      warning("Since argument method!=\"ssgsea\", genes with constant expression values are discarded.")

+      expr <- expr[sdGenes > 0 & !is.na(sdGenes), ]

+    }

+  }

+  

+  expr

+}

\ No newline at end of file

new file mode 100644

@@ -0,0 +1,15 @@

+.filterFeatures <- function(expr, method) {

+

+  ## filter out genes with constant expression values

+  sdGenes <- apply(expr, 1, sd)

+  if (any(sdGenes == 0) || any(is.na(sdGenes))) {

+    warning(sum(sdGenes == 0 | is.na(sdGenes)),

+            " genes with constant expression values throuhgout the samples.")

+    if (method != "ssgsea") {

+      warning("Since argument method!=\"ssgsea\", genes with constant expression values are discarded.")

+      expr <- expr[sdGenes > 0 & !is.na(sdGenes), ]

+    }

+  } 

+

+  expr

+}