Merge pull request #13 from CCBR/iss-11

feat: create DESeq object from RSEM expected counts

DESCRIPTION

@@ -16,6 +16,8 @@ BugReports: https://github.com/CCBR/reneeTools/issues
 Depends:
     R (>= 2.10)
 Imports:
+    assertthat,
+    DESeq2,
     dplyr,
     tidyr
 Suggests:

NAMESPACE

@@ -1,6 +1,7 @@
 # Generated by roxygen2: do not edit by hand
 
 export("%>%")
+export(create_deseq_obj)
 export(filter_low_counts)
 export(read_raw_counts)
 importFrom(dplyr,"%>%")

NEWS.md

@@ -5,3 +5,4 @@
 ## New functions
 
 - `filter_low_counts()` (#10)
+- `create_deseq_obj()` (#13)

R/data.R

@@ -1,7 +1,7 @@
-#' RSEM gene counts
+#' RSEM expected gene counts
 #'
 #' @format ## `gene_counts`
-#' A data frame with columns 'gene_id', 'GeneName', and a column for each sample's count.
+#' A data frame with columns 'gene_id', 'GeneName', and a column for each sample's expected count.
 #'
 #' @source Generated by running RENEE v2.5.8 on the
 #' [test dataset](https://github.com/CCBR/RENEE/tree/e08f7db6c6e638cfd330caa182f64665d2ef37fa/.tests)

R/deseq2.R

@@ -0,0 +1,37 @@
+#' Create DESeq2 object from gene counts and sample metadata
+#'
+#' @param counts_tbl expected gene counts from RSEM as a data frame or tibble.
+#' @param meta_dat   sample metadata as a data frame with rownames as sample IDs.
+#' @param design     model formula for experimental design. Columns must exist in `meta_dat`.
+#'
+#' @return DESeqDataSet
+#' @export
+#'
+#' @examples
+#' sample_meta <- data.frame(
+#'   row.names = c("KO_S3", "KO_S4", "WT_S1", "WT_S2"),
+#'   condition = factor(c("knockout", "knockout", "wildtype", "wildtype"),
+#'     levels = c("wildtype", "knockout")
+#'   )
+#' )
+#' dds <- create_deseq_obj(gene_counts, sample_meta, ~condition)
+create_deseq_obj <- function(counts_tbl, meta_dat, design) {
+  gene_id <- GeneName <- NULL
+  counts_dat <- counts_tbl %>%
+    # deseq2 requires integer counts
+    dplyr::mutate(dplyr::across(
+      dplyr::where(is.numeric),
+      \(x) as.integer(round(x, 0))
+    )) %>%
+    as.data.frame()
+  row.names(counts_dat) <- counts_dat %>% dplyr::pull("gene_id")
+  # convert counts tibble to matrix
+  counts_mat <- counts_dat %>%
+    dplyr::select(-c(gene_id, GeneName)) %>%
+    as.matrix()
+
+  # sample IDs must be in the same order
+  assertthat::are_equal(colnames(counts_mat), rownames(meta_dat))
+
+  return(DESeq2::DESeqDataSetFromMatrix(counts_mat, meta_dat, design))
+}

R/filter_low_counts.R

@@ -1,6 +1,6 @@
 #' filter_low_counts
 #'
-#' @param counts_dat dataframe of expected gene counts from RSEM
+#' @param counts_dat expected gene counts from RSEM as a data frame or tibble
 #' @param min_counts integer number of minimum counts across all samples (default: 0)
 #'
 #' @return filtered counts dataframe

tests/testthat/test-deseq2.R

@@ -0,0 +1,48 @@
+set.seed(20231228)
+test_that("create_deseq_obj works", {
+  sample_meta <-
+    data.frame(
+      row.names = c("KO_S3", "KO_S4", "WT_S1", "WT_S2"),
+      condition = factor(
+        c("knockout", "knockout", "wildtype", "wildtype"),
+        levels = c("wildtype", "knockout")
+      )
+    )
+  dds <- create_deseq_obj(gene_counts, sample_meta, ~condition)
+
+  expect_equal(
+    dds@colData %>% as.data.frame(),
+    structure(
+      list(condition = structure(
+        c(2L, 2L, 1L, 1L),
+        levels = c(
+          "wildtype",
+          "knockout"
+        ),
+        class = "factor"
+      )),
+      class = "data.frame",
+      row.names = c(
+        "KO_S3",
+        "KO_S4", "WT_S1", "WT_S2"
+      )
+    )
+  )
+  expect_equal(
+    dds@assays@data@listData %>% as.data.frame() %>% dplyr::filter(counts.KO_S3 > 15),
+    structure(
+      list(
+        counts.KO_S3 = c(25L, 16L, 19L),
+        counts.KO_S4 = c(22L, 10L, 26L),
+        counts.WT_S1 = c(74L, 0L, 10L),
+        counts.WT_S2 = c(104L, 0L, 8L)
+      ),
+      class = "data.frame",
+      row.names = c(
+        "ENSG00000185658.13",
+        "ENSG00000233922.2",
+        "ENSG00000157601.14"
+      )
+    )
+  )
+})