diff --git a/dev/download.R b/dev/download.R
index b0f7204..0ad9d7d 100644
--- a/dev/download.R
+++ b/dev/download.R
@@ -16,8 +16,7 @@ library(HDF5Array)
 db <- db()
 
 # Arguments
-args = commandArgs(trailingOnly=TRUE)
-root_directory = args[[1]]
+root_directory = "/vast/projects/cellxgene_curated/raw_data_Apr_2024"
 
 files_metadata = 
 	datasets(db) |>
@@ -28,6 +27,30 @@ files_metadata =
 
 files_metadata |> saveRDS(glue("{root_directory}/files_metadata.rds"))
 
+
+test = 
+  files_metadata |> 
+  slice(1:50) |> 
+  nest(data = c(dataset_id, dataset_version_id, filetype, url)) |> 
+  mutate(has_donor_id = map_lgl(
+    data,
+    ~ {
+      h5_path = .x |> files_download(dry.run = FALSE)
+      has_donor_id = 
+        h5_path |> 
+        readH5AD(use_hdf5 = TRUE	) |> 
+        colData() |> 
+        as_tibble() |> 
+        select(any_of("donor_id")) |> 
+        ncol() >
+        0
+      file.remove(h5_path)
+      has_donor_id
+    }
+  )) |> 
+  unnest(data) |> 
+  select(dataset_version_id, has_donor_id)
+
 files_metadata |>
 
 	# Get organism list and filter human
diff --git a/dev/get_pseudobulk.R b/dev/get_pseudobulk.R
deleted file mode 100644
index d23ec84..0000000
--- a/dev/get_pseudobulk.R
+++ /dev/null
@@ -1,102 +0,0 @@
-library(dplyr)
-library(cellxgenedp)
-library(tidyverse)
-#library(tidySingleCellExperiment)
-library(stringr)
-library(scMerge)
-library(glue)
-library(DelayedArray)
-library(HDF5Array)
-library(openssl)
-library(stringr)
-library(CuratedAtlasQueryR)
-library(purrr)
-
-library(dbplyr)
-library(DBI)
-library(duckdb)
-library(tidySingleCellExperiment)
-library(tidySummarizedExperiment)
-library(magrittr)
-
-metadata_DB = "/vast/projects/cellxgene_curated/metadata_annotated_0.2.3.parquet"
-root_directory = "/vast/projects/cellxgene_curated"
-
-metadata =
-  duckdb() |>
-  dbConnect(drv = _, read_only = TRUE) |>
-  tbl(metadata_DB) |>
-
-  filter(confidence_class < 4) |>
-
-  # Attach lineage
-  left_join(
-    read_csv("~/PostDoc/immuneHealthyBodyMap/metadata_cell_type.csv") |>
-      replace_na(list(lineage_1 = "other_non_immune")),
-    copy=TRUE
-  ) |>
-  mutate(is_immune = if_else(lineage_1 == "immune", "immune", "other")) |>
-
-  # Fix typo
-  mutate(tissue_harmonised = tissue_harmonised |> str_replace("plcenta", "placenta"))
-
-
-# # CREATE MAKEFILE
-# data_directory = glue("{root_directory}/0.2.1/original")
-# output_directory = glue("/home/users/allstaff/mangiola.s/PostDoc/immuneHealthyBodyMap/pseudobulk_0.2.3.4")
-# script_directory = "/home/users/allstaff/mangiola.s/PostDoc/CuratedAtlasQueryR/dev"
-# 
-# output_directory |> dir.create( showWarnings = FALSE, recursive = TRUE)
-# tab = "\t"
-# metadata |>
-#   distinct(file_id, tissue_harmonised, cell_type_harmonised, is_immune) |>
-#   as_tibble() |>
-#   mutate(
-#     tissue_harmonised = tissue_harmonised |> str_replace_all(" ", "__"),
-#     cell_type_harmonised = cell_type_harmonised |> str_replace_all(" ", "__")
-#   ) |>
-#   mutate(
-#     file_name = glue("{tissue_harmonised}____{cell_type_harmonised}____{is_immune}____{file_id}.rds") |>
-#       as.character()|>
-#       str_replace_all("/", "__"),
-#     output_file_path =
-#       glue("{output_directory}/{file_name}") |>
-#       as.character()
-#   ) |>
-# 	mutate(commands = pmap(
-# 	  list(output_file_path, file_id, tissue_harmonised, cell_type_harmonised, is_immune),
-# 		~
-# 			c(
-# 				glue("CATEGORY=pseudobulk\nMEMORY=30000\nCORES=1\nWALL_TIME=30000"),
-# 				glue(
-# 					"{..1}:{metadata_DB}\n{tab}Rscript {script_directory}/get_pseudobulk.R {..2} {..3} {..4} {..5} {..1}"
-# 				)
-# 			)
-# 	))  |>
-# 	pull(commands) |>
-# 	unlist() |>
-# 	write_lines(glue(glue("{output_directory}/get_pseudobulk.makeflow")))
-
-
-# Read arguments
-args = commandArgs(trailingOnly = TRUE)
-file_id = args[[1]]
-tissue_harmonised = args[[2]] |> str_replace_all("____", " ")
-cell_type_harmonised = args[[3]] |> str_replace_all("____", " ")
-is_immune = args[[4]] 
-output_file = args[[5]] 
-
-metadata  = 
-  metadata |>
-  filter(
-    file_id==!!file_id & 
-      tissue_harmonised == !!tissue_harmonised & 
-      cell_type_harmonised == !!cell_type_harmonised &
-      is_immune == !!is_immune
-  ) |>
-    get_single_cell_experiment(
-      cache_directory = "/vast/projects/cellxgene_curated"
-    ) |>
-    aggregate_cells(c(sample_, cell_type_harmonised)) |>
-    saveRDS(output_file)
-
diff --git a/dev/make_pseudobulk.R b/dev/make_pseudobulk.R
new file mode 100644
index 0000000..75d8a90
--- /dev/null
+++ b/dev/make_pseudobulk.R
@@ -0,0 +1,316 @@
+library(tidyverse)
+library(targets)
+library(glue)
+# Get input
+
+
+result_directory = "/vast/projects/cellxgene_curated/pseudobulk_0.2.3.4"
+
+tar_script(
+  {
+    
+    #-----------------------#
+    # Input
+    #-----------------------#
+    library(tidyverse)
+    library(targets)
+    library(tarchetypes)
+    library(glue)
+    library(qs)
+    library(crew)
+    library(crew.cluster)
+    
+    #-----------------------#
+    # Packages
+    #-----------------------#
+    tar_option_set(
+      packages = c(
+        "CuratedAtlasQueryR", "stringr", "tibble", "tidySingleCellExperiment", "dplyr", "Matrix",
+        "Seurat", "glue", "qs",  "purrr", "tidybulk", "tidySummarizedExperiment", "edgeR", "crew", "magrittr", "digest", "glmmSeq", "readr", "forcats"
+      ),
+      
+      garbage_collection = TRUE,
+      #trust_object_timestamps = TRUE,
+      memory = "transient",
+      storage = "worker",
+      retrieval = "worker",
+      #error = "continue",
+      format = "qs",
+      
+      #-----------------------#
+      # SLURM
+      #-----------------------#
+      controller = crew_controller_group(
+        
+        
+        crew_controller_slurm(
+          name = "slurm_1_5",
+          slurm_memory_gigabytes_per_cpu = 5,
+          slurm_cpus_per_task = 1,
+          workers = 200,
+          verbose = T
+        ),
+        crew_controller_slurm(
+          name = "slurm_1_10",
+          slurm_memory_gigabytes_per_cpu = 10,
+          slurm_cpus_per_task = 1,
+          workers = 100,
+          verbose = T
+        ),
+        crew_controller_slurm(
+          name = "slurm_1_20",
+          slurm_memory_gigabytes_per_cpu = 20,
+          slurm_cpus_per_task = 1,
+          workers = 50,
+          verbose = T
+        ),
+        crew_controller_slurm(
+          name = "slurm_1_40",
+          slurm_memory_gigabytes_per_cpu = 40,
+          slurm_cpus_per_task = 1,
+          workers = 30,
+          verbose = T
+        ),
+        crew_controller_slurm(
+          name = "slurm_1_80",
+          slurm_memory_gigabytes_per_cpu = 80,
+          slurm_cpus_per_task = 1,
+          workers = 20,
+          verbose = T
+        ),
+        crew_controller_slurm(
+          name = "slurm_2_400",
+          slurm_memory_gigabytes_per_cpu = 400,
+          slurm_cpus_per_task = 2,
+          workers = 1,
+          verbose = T
+        )
+      ),
+      resources = tar_resources(crew = tar_resources_crew("slurm_2_20")) ,
+      debug = "pseudobulk_file_id_3e12a178"
+      #, # Set the target you want to debug.
+      #cue = tar_cue(mode = "never")		
+    )
+    
+    split_metadata = function(){
+      
+      CAQ_directory = "/vast/projects/cellxgene_curated"
+      
+      my_metadata =
+        get_metadata(cache_directory = CAQ_directory) |>
+        distinct(file_id, sample_, cell_, file_id_db, cell_type_harmonised) |>
+        as_tibble() |> 
+        mutate(chunk = sample_) |> 
+        nest(data = -c(chunk, file_id)) |> 
+        mutate(number_of_cells = map_int(data, nrow))
+      
+    }
+    
+    get_sce = 	function(tissue_cell_type_metadata) {
+      
+      tissue_cell_type_metadata |>
+        mutate(data = pmap(
+          list(data),
+          ~ ..1 |>
+            get_single_cell_experiment(cache_directory = "/vast/projects/cellxgene_curated") 
+          # |>
+          # 	mutate(sample_se =
+          # 				 	
+          # 				 	# I need to fix Curated CellAtlas with disease sample, duplication for 
+          # 				 	# file_id=="cc3ff54f-7587-49ea-b197-1515b6d98c4c", cell_type_harmonised=="stromal_cell"
+          # 				 	# for lung
+          # 				 	glue("{sample_}___{disease}") |>
+          # 				 	str_replace_all(" ", "_") |>
+          # 				 	str_replace_all("/", "__")
+          # 	) 
+        ))
+      
+    }
+    
+    get_pseudobulk = 	function(sce_df) {
+      
+      sce_df |>
+        mutate(data = map(
+          data, 
+          ~ {
+            .x = tidySingleCellExperiment::aggregate_cells(.x, .sample = c(sample_, cell_type_harmonised)	)
+            
+            # Decrease size
+            # We will reattach rowames later
+            my_assay = assay(.x, "counts") |>  as("sparseMatrix")
+            rownames(my_assay) = NULL
+            SummarizedExperiment(assays = my_assay, colData = colData(.x))
+            
+          }
+        ))
+      
+    }
+    
+    aggregate = function(se_df, se_rownames){
+      
+      print("Start aggregate")
+      gc()
+      
+      se = 
+        se_df |>
+        
+        # Add columns and filter
+        mutate(data = map2(
+          data, file_id,
+          ~ {
+            
+            ## TEMPORARY FIX BECAUSE I FORGOT TO ADD ASSAY NAME
+            assays(.x) = .x |> assays() |> as.list() |>  set_names("counts")
+            #####
+            
+            # Add columns
+            se = 
+              .x |>
+              mutate(file_id = .y) |>
+              select(-any_of(c("file_id_db", ".cell", "original_cell_id")))
+            
+            
+            # Identify samples with many genes
+            sample_with_many_genes =
+              se |>
+              assay("counts") |>
+              apply(2, function(x) (x>0) |> which() |> length()) |>
+              enframe() |>
+              mutate(name = as.character(name)) |>
+              filter(value > 5000) |>
+              pull(name)
+            se = se[,sample_with_many_genes, drop=FALSE]
+            
+            # Filter samples with too few cells
+            se = se |> filter(.aggregated_cells > 10)
+            
+            se
+            
+          }, 
+          .progress=TRUE
+        )) |>
+        
+        nest(data = -file_id) |> 
+        mutate(data = map(data,
+                          ~ {
+                            se = .x |> 
+                              pull(data) |> 
+                              do.call(cbind, args= _)
+                            
+                            # # Filter very rare gene-transcripts
+                            # all_zeros = assay(se, "counts") |> rowSums() |> equals(0) 
+                            # se = se[!all_zeros,]
+                            # lower_than_total_counts = assay(se, "counts") |> rowSums() < 15 
+                            # se = se[!lower_than_total_counts,]
+                            
+                            
+                            # Make it sparse
+                            se@assays@data$counts = se@assays@data$counts |> as("sparseMatrix")
+                            
+                            # Attach rownames back
+                            rownames(se) = se_rownames
+                            
+                            se
+                          }))
+      
+    }
+    
+    
+    #-----------------------#
+    # Pipeline
+    #-----------------------#
+    list(
+      
+      # # Do metadata
+      # tarchetypes::tar_group_by(
+      #   chunks_df,
+      #   define_chunks(),
+      #   file_id, sample_chunk, 
+      #   resources = tar_resources(crew = tar_resources_crew("slurm_1_40"))
+      # ),
+      
+      # Get rownames
+      tar_target(
+        sce_rownames,
+        get_metadata() |> head(1) |> get_single_cell_experiment() |> rownames(),
+        resources = tar_resources(crew = tar_resources_crew("slurm_1_20"))
+      ),
+      
+      # Do metadata
+      tar_target(
+        metadata_split,
+        split_metadata(),
+        resources = tar_resources(crew = tar_resources_crew("slurm_1_40"))
+      ),
+      
+      # Get SCE SMALL
+      tarchetypes::tar_group_by(
+        metadata_split_SMALL,
+        metadata_split[metadata_split$number_of_cells<2500,],
+        chunk,
+        resources = tar_resources(crew = tar_resources_crew("slurm_1_20"))
+      ),
+      
+      tarchetypes::tar_group_by(
+        metadata_split_BIG,
+        metadata_split[metadata_split$number_of_cells>=2500,],
+        chunk,
+        resources = tar_resources(crew = tar_resources_crew("slurm_1_20"))
+      ),
+      
+      # Get SCE SMALL
+      tar_target(
+        metadata_split_pseudobulk_SMALL,
+        metadata_split_SMALL |>  get_sce() |> get_pseudobulk(),
+        pattern = map(metadata_split_SMALL),
+        iteration = "group",
+        resources = tar_resources(crew = tar_resources_crew("slurm_1_20"))
+      ),
+      
+      # Get SCE SMALL
+      tar_target(
+        metadata_split_pseudobulk_BIG,
+        metadata_split_BIG |>  get_sce() |> get_pseudobulk(),
+        pattern = map(metadata_split_BIG),
+        iteration = "group",
+        resources = tar_resources(crew = tar_resources_crew("slurm_1_80"))
+      ),
+      
+      # Group samples
+      tarchetypes::tar_group_by(
+        metadata_grouped_pseudobulk,
+        metadata_split_pseudobulk_SMALL |> bind_rows(metadata_split_pseudobulk_BIG),
+        file_id,
+        resources = tar_resources(crew = tar_resources_crew("slurm_2_400")), 
+        deployment = "main"
+      ),
+      
+      # Aggregate
+      tar_target(
+        pseudobulk_file_id,
+        metadata_grouped_pseudobulk |> aggregate(sce_rownames) ,
+        pattern = map(metadata_grouped_pseudobulk),
+        iteration = "group",
+        resources = tar_resources(crew = tar_resources_crew("slurm_1_20"))
+      )
+      
+    )
+    
+    
+  }, 
+  ask = FALSE, 
+  script = glue("{result_directory}/_targets.R")
+)
+
+
+tar_make(
+  #callr_function = NULL,
+  # reporter = "summary",
+  script = glue("{result_directory}/_targets.R"),
+  store = result_directory
+)
+
+
+
+# tar_read(pseudobulk_file_id, store = "/vast/projects/cellxgene_curated/pseudobulk_0.2.3.4")
+