From e3c8659c8580bef5d96214f87bb65dbd6944eaa2 Mon Sep 17 00:00:00 2001
From: Pablo Moreno <pcm32@users.noreply.github.com>
Date: Thu, 30 Mar 2023 23:04:50 +0100
Subject: [PATCH] AnnData Ops: Adds cell metadata merging (#289)

* Adds cell metadata merging

* Avoids pandas object dtype and adds passing tests
---
 .../scanpy/anndata_operations.xml             | 63 ++++++++++++++++++-
 .../test-data/test_incomplete_metadata.tsv    |  6 ++
 2 files changed, 67 insertions(+), 2 deletions(-)
 create mode 100644 tools/tertiary-analysis/scanpy/test-data/test_incomplete_metadata.tsv

diff --git a/tools/tertiary-analysis/scanpy/anndata_operations.xml b/tools/tertiary-analysis/scanpy/anndata_operations.xml
index 0414b35f..0bb1f16d 100644
--- a/tools/tertiary-analysis/scanpy/anndata_operations.xml
+++ b/tools/tertiary-analysis/scanpy/anndata_operations.xml
@@ -1,5 +1,5 @@
 <?xml version="1.0" encoding="utf-8"?>
-<tool id="anndata_ops" name="AnnData Operations" version="@TOOL_VERSION@+galaxy9" profile="@PROFILE@">
+<tool id="anndata_ops" name="AnnData Operations" version="@TOOL_VERSION@+galaxy91" profile="@PROFILE@">
   <description>modifies metadata and flags genes</description>
   <macros>
     <import>scanpy_macros2.xml</import>
@@ -41,6 +41,10 @@ ln -s '${copy_r.r_source}' r_source.h5 &&
   ln -s '${us}' uns_source_${i}.h5 &&
 #end for
 #end if
+
+#if $add_cell_metadata.default:
+  ln -s ${add_cell_metadata.file} cell_metadata.tsv &&
+#end if
 python $operations
 ]]></command>
   <configfiles>
@@ -57,9 +61,45 @@ def make_column_values_unique(df, field, new_field=None, suffix = '-duplicate-')
   appendents = (suffix + df.groupby(field).cumcount().astype(str).replace('0','')).replace(suffix, '')
   df[new_field] = df[field].astype(str) + appendents.astype(str)
   return df
-
+	    
 adata = sc.read('input.h5')
 
+#if $add_cell_metadata.default:
+import pandas as pd
+
+def add_cell_metadata(ad, metadata_file="cell_metadata.tsv", drop_duplicates=True):
+  metadata_df = pd.read_csv(metadata_file, sep="\t", index_col=0)
+  # we avoid renames in the original object or outright drop the column in the metadata
+  for col in ad.obs.columns:
+      if col in metadata_df.columns:
+          print(f"Renaming {col} to {col}_x")
+          if drop_duplicates:
+              metadata_df = metadata_df.drop(col, axis=1)
+          else:
+              metadata_df.rename(columns={col: col + "_x"}, inplace=True)
+  # merge metadata into ad.obs column by column, changing columns to category dtype if they become object dtype on merge
+  merged_obs = ad.obs.merge(
+      metadata_df, left_index=True, right_index=True, how="left"
+  )
+  for o_col in metadata_df.columns:
+      col = o_col
+      # lets consider cases where columns where renamed during merge
+      if o_col + "_x" in merged_obs.columns:
+          col = o_col + "_x"
+      if o_col + "_y" in merged_obs.columns:
+          col = o_col + "_y"
+      if col in merged_obs.columns:
+          if merged_obs[col].dtype == object:
+              prev_dtype = metadata_df[o_col].dtype
+              if prev_dtype == str or prev_dtype == object:
+                  prev_dtype = "category"
+              print(f"Changing {col} from {merged_obs[col].dtype} to {prev_dtype}")
+              merged_obs[col] = merged_obs[col].astype(prev_dtype)
+  return merged_obs
+	    
+adata.obs = add_cell_metadata(adata)
+#end if
+
 #if $copy_adata_to_raw:
 adata.raw = adata
 #end if
@@ -253,6 +293,13 @@ adata.write('output.h5', compression='gzip')
   <inputs>
     <param name="input_obj_file" argument="input-object-file" type="data" format="h5,h5ad" label="Input object in hdf5 AnnData format"/>
     <expand macro="output_object_params_no_loom"/>
+    <conditional name="add_cell_metadata">
+      <param name="default" type="boolean" checked="false" label="Merge additional cell metadata"/>
+      <when value="true">
+        <param name="file" type="data" label="Cell metadata with headers" help="A tabular file with headers, where the first column contains cell barcodes. Will be merged via a left join, so not all cells in the obs need to be in the metadata. Currently duplicated column headers will be ignored and the originals in the AnnData will be kept." format="tsv,tabular"/>
+      </when>
+      <when value="false"/>
+    </conditional>
     <param name="copy_adata_to_raw" type="boolean" label="Copy AnnData to .raw" help="If activated, it will do 'adata.raw = adata'" checked="false"/>
     <repeat name="modifications" title="Change field names in AnnData observations" min="0">
       <param name="from_obs" type="text" label="Original name" help="Name in observations that you want to change">
@@ -361,6 +408,18 @@ adata.write('output.h5', compression='gzip')
         </assert_contents>
       </output>
     </test>
+    <test>
+      <param name="input_obj_file" value="anndata_ops.h5"/>
+      <conditional name="add_cell_metadata">
+        <param name="default" value="true"/>
+        <param name="file" value="test_incomplete_metadata.tsv"/>
+      </conditional>
+      <output name="output_h5ad" ftype="h5ad">
+        <assert_contents>
+          <has_h5_keys keys="obs/cell_type"/>
+        </assert_contents>
+      </output>
+    </test>
     <test>
       <param name="input_obj_file" value="anndata_ops.h5"/>
       <repeat name="var_modifications" >
diff --git a/tools/tertiary-analysis/scanpy/test-data/test_incomplete_metadata.tsv b/tools/tertiary-analysis/scanpy/test-data/test_incomplete_metadata.tsv
new file mode 100644
index 00000000..6836810f
--- /dev/null
+++ b/tools/tertiary-analysis/scanpy/test-data/test_incomplete_metadata.tsv
@@ -0,0 +1,6 @@
+barcode	cell_type
+ERR2146881	A
+ERR2146882	A
+ERR2146883	B
+ERR2146884	B
+ERR2146885	C