Add files via upload

experiments/process_data.R

@@ -1 +1,157 @@
-
+library(Matrix)
+
+load_st_her2p <- function(st_id)
+{
+  data_dir_st <- "../Andersson_etal_2021"
+
+  st_counts <- read.csv(sprintf("%s/ST-cnts/%s.tsv.gz", data_dir_st, st_id), 
+                        sep = "\t", row.names = 1)
+  st_counts <- as(t(as.matrix(st_counts)), "Matrix")
+
+  st_meta <- read.csv(sprintf("%s/ST-spotfiles/%s_selection.tsv", data_dir_st, st_id), 
+                      sep = "\t")
+  rownames(st_meta) <- sprintf("%dx%d", st_meta$x, st_meta$y)
+
+  pat_file <- sprintf("%s/ST-pat/lbl/%s_labeled_coordinates.tsv", data_dir_st, st_id)
+
+  if(file.exists(pat_file))
+  {
+    pat <- read.csv(pat_file, sep = "\t")
+
+    pat <- pat[which(!is.na(pat$x)), ]
+    pat[, c("x2", "y2")] <- round(pat[, c("x", "y")])
+    rownames(pat) <- sprintf("%dx%d", pat$x2, pat$y2)
+
+    pat <- pat[rownames(st_meta), ]
+
+    st_meta$pat_label <- pat$label
+  }
+
+  common_spots <- intersect(rownames(st_meta), colnames(st_counts))
+  st_meta <- st_meta[common_spots, ]
+  st_counts <- st_counts[, common_spots]
+
+  return(list(counts = st_counts, meta = st_meta))
+}
+
+load_vis_tnbc <- function(vis_id)
+{
+  data_dir_st <- "../Wu_etal_2021_vis"
+
+  vis_counts <- sprintf("%s/filtered_count_matrices/%s_filtered_count_matrix/", 
+                        data_dir_st, vis_id)
+  vis_counts <- Seurat::Read10X(vis_counts, gene.column = 1)
+
+  vis_meta <- read.csv(sprintf("%s/metadata/%s_metadata_processed.csv", 
+                               data_dir_st, vis_id), row.names = 1)
+
+  vis_count <- vis_count[, rownames(vis_meta)]
+
+  return(list(counts = vis_count, meta = vis_meta))
+}
+
+load_sc_wu <- function(subtype = "HER2+")
+{
+  data_dir_sc <- "../Wu_etal_2021_sc"
+
+  sc_meta <- read.csv(sprintf("%s/metadata.csv", data_dir_sc), row.names = 1)
+
+  if(subtype != 'all')
+  {
+    sc_meta <- sc_meta[which(sc_meta$subtype == subtype), ]
+  }
+
+  sc_count <- Seurat::Read10X(data_dir_sc, gene.column = 1)
+
+  sc_count <- sc_count[, rownames(sc_meta)]
+
+  return(list(counts = sc_count, meta = sc_meta))
+}
+
+load_libd_sample <- function(sample_name)
+{
+  libd_data_dir <- "../LIBD"
+
+  layers <- read.csv(sprintf("%s/barcode_level_layer_map.tsv", libd_data_dir), sep = "\t", header = F)
+  colnames(layers) <- c("spot_name", "sample_name", "layer")
+
+  sample_counts <- Seurat::Read10X_h5(sprintf("%s/%s/filtered_feature_bc_matrix.h5", libd_data_dir, sample_name))
+
+  sample_layers <- layers[which(layers$sample_name == sample_name), ]
+  rownames(sample_layers) <- sample_layers$spot_name
+
+  loc <- read.csv(sprintf("%s/%s/tissue_positions_list.txt", libd_data_dir, sample_name), 
+                  header = F, row.names = 1)
+  loc <- loc[which(loc$V2 == 1),  ]
+  loc <- setNames(loc[, 2:3], c("Row", "Col"))
+
+  spots <- intersect(rownames(loc), rownames(sample_layers))
+  loc <- cbind(loc[spots, ], sample_layers[spots, "layer", drop = FALSE])
+  loc$sample <- sample_name
+  sample_counts <- sample_counts[, spots]
+  rs <- rowSums(sample_counts)
+  sample_counts <- sample_counts[which(rs > 0), ]
+
+  colnames(sample_counts) <- rownames(loc) <- sprintf("%s_%s", sample_name, spots)
+
+  return(list(counts = sample_counts, spatial = loc))
+}
+
+load_libd_batch <- function(sample_names)
+{
+  dt <- c()
+  for(sample_name in sample_names)
+  {
+    sample_data <- load_libd_sample(sample_name)
+
+    if(is.null(dt))
+    {
+      dt <- sample_data
+    }else
+    {
+      cg <- intersect(rownames(dt$counts), rownames(sample_data$counts))
+      dt$counts <- cbind(dt$counts[cg, ], sample_data$counts[cg, ])
+      dt$spatial <- rbind(dt$spatial, sample_data$spatial)
+    }
+  }
+
+  return(dt)
+}
+
+load_libd_experiment <- function(ref, srt)
+{
+  libd_data_dir <- "../LIBD"
+
+  mtdata <- read.csv(sprintf("%s/libd_clinicaldata.csv", libd_data_dir))
+  rownames(mtdata) <-  mtdata$SlideID <- as.character(mtdata$SlideID)
+
+  sample_names <- mtdata$SlideID
+
+  srt_samples <- c()
+  if(srt %in% sample_names)
+  {
+    srt_samples <- srt
+  }else
+  {
+    suppressWarnings(srt_samples <- sample_names[as.integer(srt)])
+  }
+
+  ref_samples <- c()
+  if(ref == 'aggregated')
+  {
+    ref_samples <- sample_names
+  }else if(ref == 'brain')
+  {
+    ref_samples <- sample_names[which(mtdata$Brain.Number == mtdata[srt_samples, "Brain.Number"])]
+  }else if(ref == 'adjacent')
+  {
+    ref_samples <- sample_names[which(mtdata$Brain.Number == mtdata[srt_samples, "Brain.Number"] &
+                                        mtdata$Position == mtdata[srt_samples, "Position"])]
+  }else{
+    stop("Invalid ref.")
+  }
+
+    ref_samples <- setdiff(ref_samples, srt_samples)
+  return(list(ref = load_libd_batch(ref_samples), 
+              srt = load_libd_batch(srt_samples)))
+}

experiments/run_her2p.R

@@ -0,0 +1,11 @@
+source("process_data.R")
+
+st_id <- "G2"
+sc_data <- load_sc_wu('HER2+')
+st_data <- load_st_her2p(st_id)
+
+dot_st <- DOT::setup.srt(st_data$counts, st_data$meta[, c("x", "y")], radius = sqrt(2)+0.1)
+dot_sc <- DOT::setup.ref(sc_data$counts, sc_data$meta[, "celltype_major"], max_genes = 10000)
+
+dot <- DOT::create.DOT(dot_st, dot_sc)
+dot <- DOT::run.DOT.lowresolution(dot, ratios_weight = 0.25, max_size = 200)

experiments/run_libd.R

@@ -0,0 +1,14 @@
+source('process_data.R')
+
+ref_type <- "brain"
+vis_index <- "1"
+
+libd_data <- load_libd_experiment(ref_type, vis_index)
+
+dot_ref <- DOT::setup.ref(ref_data = libd_data$ref$counts, ref_annotations = libd_data$ref$spatial$layer, max_genes = 10000)
+
+dot_srt <- DOT::setup.srt(srt_data = libd_data$srt$counts, srt_coords = libd_data$srt$spatial[, c('Row', 'Col')])
+
+dot <- DOT::create.DOT(dot_srt, dot_ref)
+
+dot <- DOT::run.DOT.lowresolution(dot, ratios_weight = ifelse(ref_type == 'aggregated', 0.25, 1), max_size = 20)

experiments/run_tnbc.R

@@ -0,0 +1,11 @@
+source("process_data.R")
+
+vis_id <- "1142243F"
+sc_data <- load_sc_wu('TNBC')
+vis_data <- load_vis_tnbc(vis_id)
+
+dot_vis <- DOT::setup.srt(vis_data$counts, vis_data$meta[, c("Col", "Row")], radius = sqrt(2)+0.1)
+dot_sc <- DOT::setup.ref(sc_data$counts, sc_data$meta[, "celltype_major"], max_genes = 10000)
+
+dot <- DOT::create.DOT(dot_vis, dot_sc)
+dot <- DOT::run.DOT.lowresolution(dot, ratios_weight = 0.25, max_size = 20)