minor improvements

src/helpers/utils.R

@@ -185,15 +185,15 @@ generate_sim_spatialCluster <- function(real_adata, sim_adata){
     metadata = sim_adata$obsm
     ))
 
-  sim_sce <- BayesSpace::spatialPreprocess(sim_sce, platform="ST", n.HVGs=2000, log.normalize=FALSE)
+  sim_sce <- BayesSpace::spatialPreprocess(sim_sce, n.PCs=7, platform="ST", n.HVGs=2000, log.normalize=FALSE)
 
   sim_sce <- BayesSpace::spatialCluster(sim_sce,
     q=max(unique(real_adata$obs[,c("spatial_cluster")])), 
-    platform="ST", 
-    d=7,
-    init.method="mclust", model="t", gamma=2,
-    nrep=1000, burn.in=100,
-    save.chain=TRUE
+    platform = "ST", 
+    d = 7,
+    init.method = "mclust", model = "t", gamma = 2,
+    nrep = 1000, burn.in = 100,
+    save.chain = TRUE
   )
   sim_cluster <- sim_sce$spatial.cluster
   return(sim_cluster)

src/metrics/ks_statistic_sc_features/script.R

@@ -8,17 +8,16 @@ requireNamespace("CARD", quietly = TRUE)
 requireNamespace("SingleCellExperiment", quietly = TRUE)
 requireNamespace("SummarizedExperiment", quietly = TRUE)
 
-
 ## VIASH START
 par <- list(
-  input_spatial_dataset = "resources_test/spatialsimbench_mobnew/MOBNEW.rds",
-  input_singlecell_dataset = "resources_test/spatialsimbench_mobnew/MOBNEW_sc.rds",
+  input_spatial_dataset = "resources_test/spatialsimbench_mobnew/dataset_sp.h5ad",
+  input_singlecell_dataset = "resources_test/spatialsimbench_mobnew/dataset_sc.h5ad",
   input_simulated_dataset = "resources_test/spatialsimbench_mobnew/simulated_dataset.h5ad",
   output = "output.h5ad"
 )
 meta <- list(
   name = "ks_statistic",
-  resources_dir = "target/executable/metrics/ks_statistics_scFeatures"
+  resources_dir = "target/executable/metrics/ks_statistic_sc_features"
 )
 ## VIASH END
 
@@ -28,26 +27,11 @@ input_real_sp <- anndata::read_h5ad(par$input_spatial_dataset)
 input_sc <- anndata::read_h5ad(par$input_singlecell_dataset)
 input_simulated_sp <- anndata::read_h5ad(par$input_simulated_dataset)
 
-
-cat("Reading input files\n")
-real_sce <- scater::logNormCounts(SingleCellExperiment::SingleCellExperiment(
-  list(counts = Matrix::t(input_real_sp$layers[["counts"]])),
-  colData = input_real_sp$obs,
-  metadata = input_real_sp$obsm
-))
-
-real_log_count <- SummarizedExperiment::assay(real_sce, "logcounts")
-real_prob_matrix <- real_sce@metadata$celltype_prop
+real_log_count <- t(input_real_sp$layers[["logcounts"]])
+real_prob_matrix <- input_real_sp$obsm[["celltype_proportions"]]
 colnames(real_prob_matrix) <- paste0("ct", seq_len(ncol(real_prob_matrix)))
-rownames(real_prob_matrix) <- colnames(real_log_count)
 
-sim_sce <- scater::logNormCounts(SingleCellExperiment::SingleCellExperiment(
-  list(counts = Matrix::t(input_simulated_sp$layers[["counts"]])),
-  colData = input_simulated_sp$obs,
-  metadata = input_simulated_sp$obsm
-))
-
-sim_log_count <- SummarizedExperiment::assay(sim_sce, "logcounts")
+sim_log_count <- scuttle::normalizeCounts(t(input_simulated_sp$layers[["counts"]]))
 
 # build cell type deconvolution in simulated data
 sim_prob_matrix <- CARD_processing(input_real_sp, input_sc)
@@ -56,19 +40,21 @@ feat_types <- c("L_stats", "celltype_interaction", "nn_correlation", "morans_I")
 
 real_scfeatures_result <- scFeatures::scFeatures(real_log_count,
   sample = rep("sample1", ncol(real_log_count)),
-  spatialCoords = list(SingleCellExperiment::colData(real_sce)$row,SingleCellExperiment::colData(real_sce)$col),
+  spatialCoords = input_real_sp$obs[c("row", "col")],
   feature_types = feat_types,
   type = "spatial_t",
   species = sc_species,
-  spotProbability =  t(real_prob_matrix))
+  spotProbability =  t(real_prob_matrix)
+)
 
 sim_scfeatures_result <- scFeatures::scFeatures(sim_log_count,
   sample = rep("sample1", ncol(sim_log_count)),
-  spatialCoords = list(SingleCellExperiment::colData(sim_sce)$row,SingleCellExperiment::colData(sim_sce)$col),
+  spatialCoords = input_simulated_sp$obs[c("row", "col")],
   feature_types = feat_types,
   type = "spatial_t",
   species = sc_species,
-  spotProbability =  t(sim_prob_matrix))
+  spotProbability =  t(sim_prob_matrix)
+)
 
 ks_statistic_L_stats <- ks::kde.test(x1 = as.numeric(real_scfeatures_result$L_stats), x2 = as.numeric(sim_scfeatures_result$L_stats))
 ks_statistic_celltype_interaction <- ks::kde.test(x1 = as.numeric(real_scfeatures_result$celltype_interaction), x2 = as.numeric(sim_scfeatures_result$celltype_interaction))

src/metrics/ks_statistic_spatial/script.py

@@ -19,17 +19,13 @@
 
 print('Reading input files', flush=True)
 input_spatial_dataset = ad.read_h5ad(par['input_spatial_dataset'])
-# input_singlecell_dataset = ad.read_h5ad(par['input_singlecell_dataset'])
 input_simulated_dataset = ad.read_h5ad(par['input_simulated_dataset'])
 
 def get_spatial_network(num_sample=None, spatial=None, radius=None, coord_type="grid", n_rings=2, set_diag=False):
     spatial_adata = ad.AnnData(np.empty((num_sample, 1), dtype="float32"))
     spatial_adata.obsm["spatial"] = spatial
-    # sq.gr.spatial_neighbors(spatial_adata, n_rings=n_rings, coord_type=coord_type, n_neighs=n_neighs, radius=radius,set_diag =set_diag)
-    sq.gr.spatial_neighbors(spatial_adata, n_rings=n_rings, coord_type=coord_type, radius=radius, set_diag=set_diag,
-                            delaunay=True)
+    sq.gr.spatial_neighbors(spatial_adata, n_rings=n_rings, coord_type=coord_type, radius=radius, set_diag=set_diag, delaunay=True)
     sn = spatial_adata.obsp["spatial_connectivities"]
-
     return sn