fix bug in perfect integration get obs

Author: ghar1821

src/control_methods/perfect_integration/script.py

@@ -4,8 +4,8 @@
 # The following code has been auto-generated by Viash.
 par = {
     "input_unintegrated": "resources_test/task_cyto_batch_integration/mouse_spleen_flow_cytometry_subset/unintegrated.h5ad",
-    "output_integrated_split1": "resources_test/task_cyto_batch_integration/mouse_spleen_flow_cytometry_subset/integrated_split1.h5ad",
-    "output_integrated_split2": "resources_test/task_cyto_batch_integration/mouse_spleen_flow_cytometry_subset/integrated_split2.h5ad",
+    "output_integrated_split1": "resources_test/task_cyto_batch_integration/mouse_spleen_flow_cytometry_subset/perfect_integrated_split1.h5ad",
+    "output_integrated_split2": "resources_test/task_cyto_batch_integration/mouse_spleen_flow_cytometry_subset/perfect_integrated_split2.h5ad",
 }
 meta = {"name": "perfect_integration"}
 

src/metrics/cms/script.R

@@ -43,11 +43,13 @@ integrated_split2 <- anndataR::read_h5ad(par[["input_integrated_split2"]])
 cat("Fetching metadata from unintegrated\n")
 integrated_split1 <- get_obs_var_for_integrated(
     i_adata = integrated_split1,
-    u_adata = unintegrated
+    u_adata = unintegrated,
+    split_id = 1
 )
 integrated_split2 <- get_obs_var_for_integrated(
     i_adata = integrated_split2,
-    u_adata = unintegrated
+    u_adata = unintegrated,
+    split_id = 2
 )
 
 # Get markers to correct

src/metrics/flowsom_mapping_similarity/script.R

@@ -6,10 +6,14 @@ par <- list(
   "input_unintegrated" = 'resources_test/task_cyto_batch_integration/mouse_spleen_flow_cytometry_subset/unintegrated.h5ad',
   "input_integrated_split1" = 'resources_test/task_cyto_batch_integration/mouse_spleen_flow_cytometry_subset/integrated_split1.h5ad',
   "input_integrated_split2" = 'resources_test/task_cyto_batch_integration/mouse_spleen_flow_cytometry_subset/integrated_split2.h5ad',
+  # if using perfect integration
+  # input_integrated_split1 = "resources_test/task_cyto_batch_integration/mouse_spleen_flow_cytometry_subset/perfect_integrated_split1.h5ad",
+  # input_integrated_split2 = "resources_test/task_cyto_batch_integration/mouse_spleen_flow_cytometry_subset/perfect_integrated_split2.h5ad",
   "output" = 'resources_test/task_cyto_batch_integration/mouse_spleen_flow_cytometry_subset/score.h5ad'
 )
 meta <- list(
-  "name" = 'flowsom_mapping_similarity'
+  "name" = 'flowsom_mapping_similarity',
+  "resources_dir" = "src/utils"
 )
 ## VIASH END
 
@@ -18,19 +22,16 @@ source(paste0(meta$resources_dir, "/helper_functions.R"))
 library(anndata)
 
 unintegrated <- anndata::read_h5ad(par[["input_unintegrated"]])
-integrated_s1 <- anndata::read_h5ad(par[["input_integrated_split1"]]) 
 
-print(unintegrated)
-print(integrated_s1)
 # read and filter split 1 data
 integrated_s1 <- anndata::read_h5ad(par[["input_integrated_split1"]]) |>
-  get_obs_var_for_integrated(unintegrated) |>
+  get_obs_var_for_integrated(unintegrated, split_id = 1) |>
   subset_nocontrols() |>
   remove_unlabelled()
 
 # read and filter split 2 data
 integrated_s2 <- anndata::read_h5ad(par[["input_integrated_split2"]]) |>
-  get_obs_var_for_integrated(unintegrated) |>
+  get_obs_var_for_integrated(unintegrated, split_id = 2) |>
   subset_nocontrols() |>
   remove_unlabelled()
 

src/utils/helper_functions.R

@@ -14,9 +14,10 @@ requireNamespace("anndataR", quietly = TRUE)
 #'
 #' @param i_adata AnnData object, integrated data
 #' @param u_adata AnnData object, unintegrated dataset
+#' @param split_id numeric, split id of the integrated data
 #' @return AnnData object with .var and .obs added
 #'
-get_obs_var_for_integrated <- function(i_adata, u_adata) {
+get_obs_var_for_integrated <- function(i_adata, u_adata, split_id) {
 
     i_adata$obs <- u_adata$obs[i_adata$obs_names, ]
     i_adata$var <- u_adata$var[i_adata$var_names, ]
@@ -25,16 +26,23 @@ get_obs_var_for_integrated <- function(i_adata, u_adata) {
     # everything is from batch 1, but some samples need to be labelled to come from batch 2
     if (i_adata$uns["method_id"] == "perfect_integration") {
         cat(
-            "Control method 'perfect_integration' detected. Changing batch labels for split 2.\n"
+            "Control method 'perfect_integration' detected. Changing batch labels.\n"
         )
         cat("Computing new batch labels\n")
         # mutate is needed as donors that are used for controls, we won't have the mapping
         i_adata_new_batch_labels <- get_batch_label_perfect_integration(
             u_adata = u_adata,
             i_adata = i_adata,
-            split_id = 1
+            split_id = split_id
         )
 
+        # safeguard
+        if (! all(i_adata_new_batch_labels$donor == i_adata$obs$donor)) {
+            stop(
+                "Donor labels do not match between new batch labels and integrated data. This should not happen!"
+            )
+        }
+
         cat("Attaching new batch labels\n")
         i_adata$obs$batch <- i_adata_new_batch_labels$new_batch_label
     }
@@ -53,12 +61,14 @@ get_obs_var_for_integrated <- function(i_adata, u_adata) {
 #' @return a dataframe with donor and new batch label
 #'
 get_batch_label_perfect_integration <- function(u_adata, i_adata, split_id) {
+    # this return which batch sample we used for a donor for a given split
     actual_donor_batch_map <- unique(
         u_adata$obs[(u_adata$obs$split == split_id), c("donor", "batch")]
     )
-    # mutate is needed as donors that are used for controls, we won't have the mapping
+    # mutate is needed as donors that are used for controls, won't have batch_new as 
+    # the split id is 0
     i_adata_new_batch_labels <- i_adata$obs[, c("donor", "batch")] %>%
-        left_join(actual_donor_batch_map, by="donor", suffix = c("_old", "_new")) %>%
+        left_join(actual_donor_batch_map, by = "donor", suffix = c("_old", "_new")) %>%
         mutate(new_batch_label = ifelse(is.na(batch_new), batch_old, batch_new)) %>%
         select(donor, new_batch_label)