Annotate Visium HD with cell2location

scripts/visium_hd_brain/0_download_visium_HD_and_metadata.sh

@@ -0,0 +1,4 @@
+# Cell type metadata
+wget https://allen-brain-cell-atlas.s3-us-west-2.amazonaws.com/metadata/WMB-10X/20231215/views/cell_metadata_with_cluster_annotation.csv
+
+curl -O https://cf.10xgenomics.com/samples/spatial-exp/3.0.0/Visium_HD_Mouse_Brain/Visium_HD_Mouse_Brain_binned_outputs.tar.gz

scripts/visium_hd_brain/2_downsample_ABA_meta.pbs → scripts/visium_hd_brain/2_downsample_ABA_metadata.pbs

@@ -7,4 +7,4 @@
 
 module load R-bundle-Bioconductor/3.18-foss-2023a-R-4.3.2
 cd $PBS_O_WORKDIR
-Rscript --vanilla 2_downsample_ABA_meta.R
+Rscript --vanilla 2_downsample_ABA_metadata.R

scripts/visium_hd_brain/3_downsample_ABA_sc.pbs

@@ -1,10 +1,11 @@
 #!/bin/bash -l
-#PBS -l nodes=1:ppn=7
+#PBS -l nodes=1:ppn=5
+#PBS -l mem=64gb
 #PBS -l walltime=01:00:00
 #PBS -e logs/
 #PBS -o logs/
 
-module load Single-cell-python-bundle/2024.0
+module load Single-cell-python-bundle/2024.02-foss-2023a
 cd $PBS_O_WORKDIR
 python 3_downsample_ABA_sc.py ${PBS_ARRAYID}
 

scripts/visium_hd_brain/4_prepare_sc_and_st.pbs

@@ -0,0 +1,10 @@
+#!/bin/bash -l
+#PBS -l nodes=1:ppn=1
+#PBS -l walltime=00:20:00
+#PBS -l mem=80gb
+#PBS -e logs/
+#PBS -o logs/
+
+module load Single-cell-python-bundle/2024.02-foss-2023a
+cd $PBS_O_WORKDIR
+python 4_prepare_sc_and_st.py

scripts/visium_hd_brain/4_prepare_sc_and_st.py

@@ -0,0 +1,58 @@
+import os
+import anndata as ad
+import pandas as pd
+
+# Define the path to the subset folder
+subset_folder = os.environ['VSC_DATA_VO_USER'] + '/ABA_data/subset/'
+output_file = os.environ['VSC_DATA_VO_USER'] + '/ABA_data/WMB-10Xv3_subset.h5ad'
+
+# List all .h5ad files in the subset folder
+subset_files = sorted([f for f in os.listdir(subset_folder) if f.endswith('.h5ad')])
+
+# Initialize an empty list to hold the AnnData objects
+adatas = []
+
+# Read each .h5ad file and append to the list
+for subset_file in subset_files:
+    file_path = os.path.join(subset_folder, subset_file)
+    adata = ad.read_h5ad(file_path)
+    adatas.append(adata)
+
+# Concatenate all AnnData objects into a single AnnData object
+combined_adata = ad.concat(adatas, join='outer', merge='same')
+
+# Download metadata
+downsampled_csv = os.environ['VSC_DATA_VO_USER'] + '/ABA_data/cell_metadata_with_cluster_annotation_downsampled.csv'
+metadata = pd.read_csv(downsampled_csv)
+
+# Ensure that the index of the metadata dataframe is the 'cell_label' column
+metadata.set_index('cell_label', inplace=True)
+
+# Identify overlapping columns
+overlapping_columns = metadata.columns.intersection(combined_adata.obs.columns)
+print(f"Overlapping columns: {overlapping_columns}")
+
+# Rename overlapping columns in the metadata dataframe
+metadata.rename(columns={col: f"{col}_metadata" for col in overlapping_columns}, inplace=True)
+
+# Align the metadata indices with the obs_names of the AnnData object
+metadata = metadata.loc[combined_adata.obs_names]
+
+# Check if all indices are aligned
+assert all(metadata.index == combined_adata.obs_names)
+
+# Add metadata columns to the AnnData object
+# Join using the indices
+combined_adata.obs = combined_adata.obs.join(metadata)
+
+# Save the combined AnnData object to a file
+combined_adata.write(output_file)
+print(f"Combined data saved to {output_file}")
+
+##### VISIUM HD #####
+import scanpy as sc
+file_path = os.environ['VSC_DATA_VO_USER'] + '/spatialnichenet/Visium_HD_Mouse_Brain/square_008um/'
+
+# Read anndata from 10x output
+adata = sc.read_10x_h5(file_path + 'filtered_feature_bc_matrix.h5')
+adata.write(os.environ['VSC_DATA_VO_USER'] + '/rds/Visium_HD_Mouse_Brain_008um.h5ad')

scripts/visium_hd_brain/5_deconvolve.pbs

@@ -0,0 +1,10 @@
+#!/bin/bash -l
+#PBS -l nodes=1:ppn=1
+#PBS -l walltime=04:00:00
+#PBS -l mem=64gb
+#PBS -e logs/
+#PBS -o logs/
+
+module load Single-cell-python-bundle/2024.02-foss-2023a
+cd $PBS_O_WORKDIR
+python 4_prepare_sc_and_st.py

scripts/visium_hd_brain/5_deconvolve.sh

@@ -0,0 +1,6 @@
+nextflow run main.nf -profile hpc \
+-params-file $VSC_DATA_VO_USER/spatial_ccc_experiments/scripts/visium_hd_brain/cell2location_params.yaml \
+--mode run_dataset \
+--sc_input $VSC_DATA_VO_USER/ABA_data/WMB-10Xv3_subset.h5ad \
+--sp_input $VSC_DATA_VO_USER/rds/Visium_HD_Mouse_Brain_008um.h5ad \
+--annot class --methods cell2location --skip_metrics --gpu

scripts/visium_hd_brain/run_cell2location.config

@@ -0,0 +1,25 @@
+params {
+  file_type = "h5ad"
+  sc_input = "/data/gent/vo/000/gvo00070/vsc43831/ABA_data/WMB-10Xv3_subset.h5ad"
+  sp_input = "/data/gent/vo/000/gvo00070/vsc43831/rds/Visium_HD_Mouse_Brain_008um.h5ad"
+  annot = "class"
+  methods = "cell2location"
+  sampleID = "donor_label"
+  verbose = true
+  skip_metrics = true
+  deconv_args = [cell2location: [build: "", fit:"-n 2"]]
+  gpu = true
+
+}
+
+process {
+  withName: buildCell2locationModel {
+    memory = { "${32 + (task.attempt * 32)} GB" }
+    time = { "2h" }
+  }
+
+  withName: fitCell2locationModel {
+    memory = { "${64 + (task.attempt * 32)} GB" }
+    time = { "${task.attempt * 4}h" }
+  }
+}