Merge pull request #52 from aifimmunology/adt-matrix

Adds automatic ADT data detection for both 10x and AIFI injected ADT matrices to `read_h5_seurat()` and `read_h5_sce()`.

To avoid confusion with previous versions that don't do this automatically, version is updated from 1.2.0 to 1.3.0.

DESCRIPTION

@@ -1,7 +1,7 @@
 Package: H5weaver
 Type: Package
 Title: Read, Rearrange, and Write HDF5 files
-Version: 1.2.0
+Version: 1.3.0
 Author: Lucas Graybuck <[email protected]>
 Maintainer: Lucas Graybuck <[email protected]>
 Description: This package provides functions for reading, writing, combining, and rearranging HDF5 files with conventions from 10x Genomics.
@@ -21,5 +21,7 @@ Suggests:
   testthat (>= 2.1.0),
   Seurat (>= 3.1.0),
   SingleCellExperiment (>= 1.8.0),
-  SummarizedExperiment (>= 1.16.1)
+  SummarizedExperiment (>= 1.16.1),
+  assertthat,
+  R.utils
 RoxygenNote: 7.1.1

R/read_rna_h5.R

@@ -204,6 +204,10 @@ read_h5_feature_meta <- function(h5_file,
 #'
 #' @return a Seurat Class object
 #' @export
+#' @examples
+#' h5_fpath <- system.file('testdata/cite_sample1.h5', package = "H5weaver")
+#' so <- read_h5_seurat(h5_fpath)
+#' # so
 read_h5_seurat <- function(h5_file,
                            target = "matrix",
                            feature_names = "name",
@@ -228,32 +232,48 @@ read_h5_seurat <- function(h5_file,
   rownames(cell_meta) <- cell_meta$barcodes
 
   rownames(cell_meta) <- cell_meta$barcodes
-
   feat_meta <- read_h5_feature_meta(h5_file,
                                     target = target)
+  rownames(feat_meta) <- make.unique(feat_meta[[feature_names]])
 
-  cite <- FALSE
+  cite_10x <- FALSE
+  cite_injected <- FALSE
 
-  # Check for CITE-seq data
+  # Check for cellranger CITE-seq data
   if("feature_type" %in% names(feat_meta)) {
     if("Antibody Capture" %in% feat_meta$feature_type) {
-      cite <- TRUE
+      cite_10x <- TRUE
     }
   }
 
-  if(cite) {
+  if(cite_10x) {
     cite_feat <- feat_meta[feat_meta$feature_type == "Antibody Capture",]
     feat_meta <- feat_meta[feat_meta$feature_type != "Antibody Capture",]
 
     cite_mat <- mat[cite_feat$id,]
     mat <- mat[feat_meta$id,]
   }
+
+  # Check for injected CITE-seq data
+  if("ADT" %in% h5ls(h5_file)$name) {
+    cite_injected <- TRUE
+  }
 
+  if(cite_injected) {
+    cite_mat <- read_h5_dgCMatrix(h5_file, "ADT", feature_names = "id")
+    colnames(cite_mat) <- cell_meta$barcodes
+    cite_feat <- read_h5_feature_meta(h5_file, target = "ADT")
+    rownames(cite_feat) <- make.unique(cite_feat[["id"]])
+  }
+
   so <- Seurat::CreateSeuratObject(counts = mat,
                                    meta.data = cell_meta,
                                    ...)
-  if(cite) {
+  so[["RNA"]] <- Seurat::AddMetaData( so[["RNA"]], feat_meta)
+
+  if(cite_10x|cite_injected) {
     so[["ADT"]] <- Seurat::CreateAssayObject(counts = cite_mat)
+    so[["ADT"]] <- Seurat::AddMetaData( so[["ADT"]], cite_feat)
   }
 
   so
@@ -272,6 +292,10 @@ read_h5_seurat <- function(h5_file,
 #'
 #' @return a SingleCellExperiment Class object
 #' @export
+#' @examples
+#' h5_fpath <- system.file('testdata/cite_sample1.h5', package = "H5weaver")
+#' test_sce <- read_h5_sce(h5_fpath)
+#' # test_sce
 read_h5_sce <- function(h5_file,
                         target = "matrix",
                         feature_names = "name",
@@ -294,22 +318,34 @@ read_h5_sce <- function(h5_file,
   feat_meta <- read_h5_feature_meta(h5_file,
                                     target = target)
 
-  cite <- FALSE
+  cite_10x <- FALSE
+  cite_injected <- FALSE
 
   # Check for CITE-seq data
   if("feature_type" %in% names(feat_meta)) {
     if("Antibody Capture" %in% feat_meta$feature_type) {
-      cite <- TRUE
+      cite_10x <- TRUE
     }
   }
 
-  if(cite) {
+  if(cite_10x) {
     cite_feat <- feat_meta[feat_meta$feature_type == "Antibody Capture",]
     feat_meta <- feat_meta[feat_meta$feature_type != "Antibody Capture",]
 
     cite_mat <- mat[cite_feat$id,]
     mat <- mat[feat_meta$id,]
   }
+
+  # Check for injected CITE-seq data
+  if("ADT" %in% h5ls(h5_file)$name) {
+    cite_injected <- TRUE
+  }
+
+  if(cite_injected) {
+    cite_mat <- read_h5_dgCMatrix(h5_file, 'ADT', feature_names = 'id')
+    cite_feat <- read_h5_feature_meta(h5_file, 'ADT')
+    colnames(cite_mat) <- cell_meta$barcodes
+  }
 
   sce <- SingleCellExperiment::SingleCellExperiment(
     assays = list(counts = mat),
@@ -318,9 +354,10 @@ read_h5_sce <- function(h5_file,
     ...
   )
 
-  if(cite) {
+  if(cite_10x|cite_injected) {
     cite_se <- SummarizedExperiment::SummarizedExperiment(
-      assays = list(counts = cite_mat)
+      assays = list(counts = cite_mat),
+      rowData = cite_feat
     )
     SingleCellExperiment::altExp(sce, "ADT") <- cite_se
   }

inst/testdata/generate_test_data_cite.R

@@ -0,0 +1,32 @@
+library(rhdf5)
+library(H5weaver)
+library(hise)
+
+chrM_genes <- fread(system.file("reference/GRCh38_10x_chrM_gene_metadata.csv.gz",
+                                package = "H5weaver"))
+h5_fid <- '1be65739-445a-41fe-89a4-52fbbba2000d'
+hise::cacheFiles(list(h5_fid))
+sample_cite_h5 <- list.files("cache/1be65739-445a-41fe-89a4-52fbbba2000d", full.names = T)
+
+## Well 1 .h5 file
+sample_list <- h5dump(sample_cite_h5)
+sample_list <- h5_list_convert_to_dgCMatrix(sample_list)
+
+set.seed(3)
+keep_genes <- c(sample(sample_list$matrix$features$name, 1000), chrM_genes$name)
+sample_keep <- match(keep_genes, sample_list$matrix$features$name)
+
+sample_list$matrix_dgCMatrix <- sample_list$matrix_dgCMatrix[sample_keep,]
+
+sample_list$matrix$features$name <- sample_list$matrix$features$name[sample_keep]
+sample_list$matrix$features$genome <- sample_list$matrix$features$genome[sample_keep]
+sample_list$matrix$features$feature_type <- sample_list$matrix$features$feature_type[sample_keep]
+sample_list$matrix$features$interval <- sample_list$matrix$features$interval[sample_keep]
+
+sample_list <- h5_list_convert_from_dgCMatrix(sample_list)
+
+write_h5_list(sample_list,
+              "cite_sample1.h5",
+              overwrite = TRUE)
+h5closeAll()
+