Merge pull request #158 from omnideconv/fix_conversion_fuction

LorenzoMerotto · web-flow · commit 6f6f66ea5dd5 · 2025-09-27T15:48:48.000+02:00
Fix conversion fuction
diff --git a/.conda/meta.yaml b/.conda/meta.yaml
@@ -2,7 +2,7 @@
 
 package:
   name: {{ name }}
-  version: "develop" 
+  version: "develop"
 
 source:
   path: ".."
@@ -68,7 +68,7 @@ test:
     - '$R -e "library(''immunedeconv'')"'
 
 about:
-  home: https://github.com/icbi-lab/immunedeconv
+  home: https://github.com/omnideconv/immunedeconv
   license: BSD_3_clause
   summary: "collection of methods for immune cell deconvolution of bulk RNA-seq samples."
   license_family: BSD
diff --git a/DESCRIPTION b/DESCRIPTION
@@ -66,4 +66,4 @@ Roxygen: list(markdown = TRUE)
 LazyData: true
 URL: https:/omnideconv.org/immunedeconv, https://github.com/omnideconv/immunedeconv
 BugReports: https://github.com/omnideconv/immunedeconv/issues
-RoxygenNote: 7.2.3
+RoxygenNote: 7.3.3
diff --git a/NAMESPACE b/NAMESPACE
@@ -8,7 +8,6 @@ export(custom_deconvolution_methods)
 export(deconvolute)
 export(deconvolute_abis)
 export(deconvolute_base_algorithm)
-export(deconvolute_base_custom)
 export(deconvolute_cibersort)
 export(deconvolute_cibersort_custom)
 export(deconvolute_consensus_tme)
@@ -22,6 +21,7 @@ export(deconvolute_mmcp_counter)
 export(deconvolute_mouse)
 export(deconvolute_quantiseq)
 export(deconvolute_seqimmucc)
+export(deconvolute_seqimmucc_custom)
 export(deconvolute_timer)
 export(deconvolute_xcell)
 export(deconvolution_methods)
@@ -35,6 +35,7 @@ export(map_cell_types)
 export(map_result_to_celltypes)
 export(reduce_mouse_cell_types)
 export(scale_to_million)
+export(seqImmuCC_LLSR)
 export(set_cibersort_binary)
 export(set_cibersort_mat)
 export(timer_available_cancers)
diff --git a/R/custom_deconvolution_methods.R b/R/custom_deconvolution_methods.R
@@ -19,7 +19,7 @@ NULL
 #' List of methods that support the use of a custom signature
 #'
 #' The available methods are
-#' `epic`, `cibersort`, `cibersort_abs`, `consensus_tme`, `base`
+#' `epic`, `cibersort`, `cibersort_abs`, `consensus_tme`, `seqimmucc`
 #'
 #' The object is a named vector. The names correspond to the display name of the method,
 #' the values to the internal name.
@@ -29,7 +29,7 @@ custom_deconvolution_methods <- c(
   "EPIC" = "epic",
   "CIBERSORT" = "cibersort",
   "ConsensusTME" = "consensus_tme",
-  "BASE" = "base"
+  "seqImmuCC" = "seqimmucc"
 )
 
 
@@ -153,25 +153,29 @@ deconvolute_consensus_tme_custom <- function(gene_expression_matrix, signature_g
 }
 
 
-
-#' Deconvolute using BASE and a custom signature matrix
+#' Deconvolute using seqImmuCC (LLSR regression) and a custom signature matrix
 #'
 #' @param gene_expression_matrix a m x n matrix with m genes and n samples. Data
 #'    should be TPM normalized and log10 scaled.
-#' @param signature_matrix a m x l matrix with m genes and l cell types. Data
-#'    should be non normalized, as the normalization wil be done in the construction
-#'    of the compendium (internal structure)
-#' @param n_permutations the number of permutations of each sample expression
-#'    to generate. These are used to normalize the results.
-#' @param log10 logical. if TRUE, log10 transforms the expression matrix.
+#' @param signature_matrix a m x l matrix with m genes and l cell types. The
+#'   matrix should contain only a subset of the genes useful for the analysis.
+#' @param ... passed to the original seqImmuCC_LLSR function
 #' @export
 #'
-deconvolute_base_custom <- function(gene_expression_matrix,
-                                    signature_matrix,
-                                    n_permutations = 100,
-                                    log10 = TRUE) {
-  new.cell.compendium <- create_base_compendium(signature_matrix)
-  results <- base_algorithm(gene_expression_matrix, new.cell.compendium, perm = n_permutations)
+deconvolute_seqimmucc_custom <- function(gene_expression_matrix,
+                                         signature_matrix,
+                                         ...) {
+  arguments <- dots_list(
+    signature = signature_matrix,
+    SampleData = gene_expression_matrix, ..., .homonyms = "last"
+  )
+
+  call <- rlang::call2(seqImmuCC_LLSR, !!!arguments)
+  results <- eval(call)
+
+
+  # results <- seqImmuCC_LLSR(signature_matrix, gene_expression_matrix, ..., .homonyms = "last")
+  results <- results[, !colnames(results) %in% c("Correlation", "RSEM")]
 
   return(t(results))
 }
diff --git a/R/immune_deconvolution_methods.R b/R/immune_deconvolution_methods.R
@@ -328,12 +328,22 @@ deconvolute_consensus_tme <- function(gene_expression_matrix,
       t, method
     )
 
+    colnames(cur.results) <- colnames(gene_expression_matrix)[cur.samples]
     list.results[[t]] <- cur.results
   }
 
-  results <- Reduce(cbind, list.results)
+  list.results <- lapply(list.results, function(x) as.data.frame(x))
 
-  colnames(results) <- colnames(gene_expression_matrix)
+  if (length(tumor.types) > 1) {
+    results <- Reduce(
+      function(x, y) merge(x, y, all = TRUE),
+      lapply(list.results, function(x) data.frame(x, rn = row.names(x)))
+    )
+
+    results <- column_to_rownames(results, "rn")
+  } else {
+    results <- list.results[[1]]
+  }
 
   return(results)
 }
diff --git a/R/mouse_deconvolution_methods.R b/R/mouse_deconvolution_methods.R
@@ -271,6 +271,7 @@ convert_human_mouse_genes <- function(gene_expression_matrix, mirror = "www",
     gene_expression_matrix <- as.data.frame(gene_expression_matrix)
     gene_expression_matrix$gene_name <- gene.names
   } else {
+    gene.names <- gene_expression_matrix
     gene_expression_matrix <- data.frame(
       "gene_name" = gene_expression_matrix,
       "X" = rep(0, length(gene_expression_matrix))
diff --git a/R/seqImmuCC_LLSR.R b/R/seqImmuCC_LLSR.R
@@ -15,7 +15,7 @@
 #' @param sig.stand logical.If TRUE, standardizes the signature matrix
 #' @param sample.scale logical. If TRUE, scales the sample expression
 #' @param log logical. If TRUE, log transforms signature and expression data
-#'
+#' @export
 #'
 seqImmuCC_LLSR <- function(signature, SampleData, w = NA, QN = T, sig.scale = F, sig.stand = T, sample.scale = T, log = T) {
   # Expression profile format standarlization
diff --git a/_pkgdown.yml b/_pkgdown.yml
@@ -40,7 +40,7 @@ reference:
       - custom_deconvolution_methods
       - deconvolute_cibersort_custom
       - deconvolute_epic_custom
-      - deconvolute_base_custom
+      - deconvolute_seqimmucc_custom
       - deconvolute_consensus_tme_custom
   - title: Cell type mapping
     desc: Map cell types and datasets to a controlled vocabulary.
diff --git a/inst/extdata/cell_type_mapping.xlsx b/inst/extdata/cell_type_mapping.xlsx
diff --git a/man/cell_type_tree.Rd b/man/cell_type_tree.Rd
diff --git a/man/custom_deconvolution_methods.Rd b/man/custom_deconvolution_methods.Rd
diff --git a/man/deconvolute_base_custom.Rd b/man/deconvolute_base_custom.Rd
diff --git a/man/deconvolute_seqimmucc_custom.Rd b/man/deconvolute_seqimmucc_custom.Rd
diff --git a/man/immunedeconv.Rd b/man/immunedeconv.Rd
diff --git a/tests/testthat/test_custom_matrices.R b/tests/testthat/test_custom_matrices.R
@@ -5,8 +5,9 @@ test_mat <- as.matrix(test_mat)
 
 
 
-test_that("BASE works with a custom signature matrix", {
-  sign_mat <- matrix(120 * runif(1500), ncol = 10)
+test_that("seqimmucc works with a custom signature matrix", {
+  test_mat <- dataset_racle$expr_mat
+  sign_mat <- matrix(120 * runif(15000), ncol = 10)
   colnames(sign_mat) <- c(
     "A", "B", "C", "D",
     "E", "F", "G", "H",
@@ -15,7 +16,7 @@ test_that("BASE works with a custom signature matrix", {
 
   rownames(sign_mat) <- sample(rownames(test_mat), nrow(sign_mat))
 
-  res <- deconvolute_base_custom(test_mat, sign_mat)
+  res <- deconvolute_seqimmucc_custom(test_mat, sign_mat)
   assert("matrix dimensions consistent", ncol(res) == ncol(test_mat))
   assert("matrix dimensions consistent", nrow(res) == ncol(sign_mat))
 })
diff --git a/tests/testthat/test_deconvolution.R b/tests/testthat/test_deconvolution.R
@@ -84,8 +84,8 @@ test_that("consensus_tme works", {
 
 
 test_that("consensus_tme with multiple indications, ordered and unordered", {
-  indications_1 <- c("blca", "blca", "brca", "brca", "brca", "brca", "chol", "chol")
-  indications_2 <- c("brca", "brca", "blca", "chol", "chol", "brca", "brca", "blca")
+  indications_1 <- c("blca", "blca", "brca", "brca", "brca", "brca", "chol", "dlbc")
+  indications_2 <- c("brca", "brca", "blca", "chol", "dlbc", "brca", "brca", "blca")
   res_1 <- deconvolute_consensus_tme(test_mat, indications = indications_1)
   res_2 <- deconvolute_consensus_tme(test_mat, indications = indications_2)
   assert("matrix dimensions consistent", ncol(res_1) == ncol(test_mat))
diff --git a/tests/testthat/test_deconvolution_mouse.R b/tests/testthat/test_deconvolution_mouse.R
@@ -39,13 +39,18 @@ test_that("generic deconvolution works for all methods", {
   })
 })
 
-# test_that("mouse gene names can be converted into their human orthologus, and vice versa", {
-#   test_mat_newGenes <- convert_human_mouse_genes(test_mat[1:1000, ], convert_to = "human")
-#   assert("matrix dimensions consistent", ncol(test_mat_newGenes) == ncol(test_mat))
-#
-#   test_mat_human <- read_tsv("bulk_mat.tsv") %>%
-#     as.data.frame() %>%
-#     tibble::column_to_rownames("gene_symbol")
-#   test_mat_newGenes <- convert_human_mouse_genes(test_mat_human[1:1000, ], convert_to = "mouse")
-#   assert("matrix dimensions consistent", ncol(test_mat_newGenes) == ncol(test_mat_human))
-# })
+test_that("mouse gene names can be converted into their human orthologus, and vice versa", {
+  test_mat_newGenes <- convert_human_mouse_genes(test_mat[1:1000, ], convert_to = "human")
+  assert("matrix dimensions consistent", ncol(test_mat_newGenes) == ncol(test_mat))
+
+  test_mat_human <- read_tsv("bulk_mat.tsv") %>%
+    as.data.frame() %>%
+    tibble::column_to_rownames("gene_symbol")
+  test_mat_newGenes <- convert_human_mouse_genes(test_mat_human[1:1000, ], convert_to = "mouse")
+  assert("matrix dimensions consistent", ncol(test_mat_newGenes) == ncol(test_mat_human))
+})
+
+test_that("mouse gene names in vector can be converted into their human orthologus, and vice versa", {
+  newGenes <- convert_human_mouse_genes(rownames(test_mat[1:1000, ]), convert_to = "human")
+  assert("gene names are in vector form", is.vector(newGenes))
+})
diff --git a/vignettes/detailed_example_mouse.Rmd b/vignettes/detailed_example_mouse.Rmd
@@ -66,7 +66,7 @@ res_mMCPcounter %>%
 
 ## Deconvolution with human-based methods
 
-Human-based methods can still be used to deconvolve mouse data through the use of orthologous genes. The function `mouse_genes_to_human` does that by retrieving the correspondent gene names with `biomaRt`. Since the gene names are retrieved from the Ensembl database, it can happen that the command has to be run with different Emsembl mirrors (see the documentation)
+Human-based methods can still be used to deconvolve mouse data through the use of orthologous genes. The function `convert_human_mouse_genes` does that by retrieving the correspondent gene names with `biomaRt`. Since the gene names are retrieved from the Ensembl database, it can happen that the command has to be run with different Emsembl mirrors (see the documentation)
 
 
 ```R
diff --git a/vignettes/immunedeconv.Rmd b/vignettes/immunedeconv.Rmd
@@ -124,26 +124,28 @@ In addition, human-based methods can be used to deconvolute mouse data through t
 to the corresponding human orthologues
 
 ```R
-gene_expression_matrix <- immunedeconv::mouse_genes_to_human(gene_expression_matrix)
+gene_expression_matrix <- immunedeconv::convert_human_mouse_genes(gene_expression_matrix, convert_to = "human")
 immunedeconv::deconvolute(gene_expression_matrix, "quantiseq")
 ```
 
 ## Deconvolution using a custom signature 
 
 Finally, certain methods can be used with custom signatures, consisting of either a signature matrix or signature genes 
-for the cell types of interest. Since the information used to deconvolute the bulk is user-provided, these functions can be 
-used for different tissues and organisms. 
+for the cell types of interest. Since the information used to deconvolute the bulk is user-provided, these functions can be used for different tissues and organisms. 
 The functions may require different input data formats, related to the requirements of each method. Please refer to their documentation. 
 The available methods are
 
 
 ```r
-base:  deconvolute_base_custom()
 cibersort norm/abs:  deconvolute_cibersort_custom()
 epic: deconvolute_epic_custom()
 consensus_tme: deconvolute_consensus_tme_custom()
+seqimmucc: deconvolute_seqimmucc_custom()
 ```
 
+Please note that using the `deconvolute_seqimmucc_custom()` with a custom signature matrix will automatically use the LLSR regression method. 
+If you wish to use the nuSVR regression method, use the `deconvolute_cibersort_custom()` function. 
+
 
 ## Example
 For this example, we use a dataset of four melanoma patients from [@EPIC2017]. 
@@ -241,8 +243,7 @@ with(cell_type_hierarchy, {
 
 For each method, each cell-type is mapped to a node in the tree.
 If you are curious, it's all defined in [this excel
-sheet](https://github.com/grst/immunedeconv/blob/master/inst/extdata/cell_type_mapping.xlsx).
-
+sheet](https://github.com/omnideconv/immunedeconv/blob/master/inst/extdata/cell_type_mapping.xlsx).
 This tree can be used to summarize scores along the tree. For instance,
 quanTIseq provides scores for regulatory and non-regulatory CD4+ T cells independently, but you are
 interested in the fraction of overall CD4+ T cells. In that case you can use
@@ -300,15 +301,11 @@ a score in arbitrary units.
 
 # FAQs
 ### Can I specify a custom signature matrix through immunedeconv?
-No, currently not. The reason is that the methods are conceptually different.
-Some are marker gene based and others deconvolution-based. CIBERSORT performs
-feature-selection on the matrix while EPIC and quanTIseq don't. EPIC uses *all*
-genes to estimate the inter-sample variance while quanTIseq uses marker genes
-only. This is also being discussed in
-[#15](https://github.com/grst/immunedeconv/issues/15).
-
-You can, however, provide custom signatures for most individual methods (see
-next question).
+Currentl, four methods (EPIC, CIBERSORT, ConsensusTME and seqImmuCC) allow users to provide their own signature matrix and/or gene set. 
+Please note that the format of the signatures might vary across methods. 
+In addition, since the embedded signatures were optimized for the methods, user-provided ones might have a worse performance. 
+If you want to know more, please see [#15](https://github.com/grst/immunedeconv/issues/15).
+
 
 ### I want to use a special feature of a method, but I cannot access it through the `deconvolute` function.
 You can access each method individually through the `deconvolute_xxx` function.
