update scripts

src/control_methods/majority_vote/script.py

@@ -1,4 +1,5 @@
 import anndata as ad
+import pandas as pd
 
 ## VIASH START
 par = {
@@ -16,11 +17,20 @@
 input_test = ad.read_h5ad(par['input_test'])
 
 print("Compute majority vote", flush=True)
-majority = input_train.obs.label.value_counts().index[0]
+label_pred = input_train.obs.label.value_counts().index[0]
 
-print("Create prediction object", flush=True)
-input_test.obs["label_pred"] = majority
+print("Create output data", flush=True)
+output = ad.AnnData(
+    obs=pd.DataFrame(
+        { 'label_pred': label_pred },
+        index=input_test.obs.index
+    ),
+    uns={
+        'method_id': meta['name'],
+        "dataset_id": input_test.uns["dataset_id"],
+        "normalization_id": input_test.uns["normalization_id"]
+    }
+)
 
-print("Write output to file", flush=True)
-input_test.uns["method_id"] = meta["functionality_name"]
-input_test.write_h5ad(par["output"], compression="gzip")
+print("Write output data", flush=True)
+output.write_h5ad(par['output'], compression="gzip")

src/control_methods/random_labels/script.py

@@ -1,5 +1,6 @@
 import anndata as ad
 import numpy as np
+import pandas as pd
 
 ## VIASH START
 par = {
@@ -21,13 +22,25 @@
 label_distribution = label_distribution / label_distribution.sum()
 
 print("Create prediction object", flush=True)
-input_test.obs["label_pred"] = np.random.choice(
+label_pred = np.random.choice(
     label_distribution.index,
     size=input_test.n_obs,
     replace=True,
     p=label_distribution
 )
 
-print("Write output to file", flush=True)
-input_test.uns["method_id"] = meta["functionality_name"]
-input_test.write_h5ad(par["output"], compression="gzip")
+print("Create output data", flush=True)
+output = ad.AnnData(
+    obs=pd.DataFrame(
+        { 'label_pred': label_pred },
+        index=input_test.obs.index
+    ),
+    uns={
+        'method_id': meta['name'],
+        "dataset_id": input_test.uns["dataset_id"],
+        "normalization_id": input_test.uns["normalization_id"]
+    }
+)
+
+print("Write output data", flush=True)
+output.write_h5ad(par['output'], compression="gzip")

src/control_methods/true_labels/script.py

@@ -1,3 +1,4 @@
+import pandas as pd
 import anndata as ad
 
 ## VIASH START
@@ -18,8 +19,20 @@
 input_solution = ad.read_h5ad(par['input_solution'])
 
 print("Create prediction object", flush=True)
-input_test.obs["label_pred"] = input_solution.obs["label"]
+label_pred = input_solution.obs["label"]
 
-print("Write output to file", flush=True)
-input_test.uns["method_id"] = meta["functionality_name"]
-input_test.write_h5ad(par["output"], compression="gzip")
+print("Create output data", flush=True)
+output = ad.AnnData(
+    obs=pd.DataFrame(
+        { 'label_pred': label_pred },
+        index=input_test.obs.index
+    ),
+    uns={
+        'method_id': meta['name'],
+        "dataset_id": input_test.uns["dataset_id"],
+        "normalization_id": input_test.uns["normalization_id"]
+    }
+)
+
+print("Write output data", flush=True)
+output.write_h5ad(par['output'], compression="gzip")

src/methods/knn/script.py

@@ -1,3 +1,4 @@
+import pandas as pd
 import anndata as ad
 import sklearn.neighbors
 
@@ -8,7 +9,7 @@
     'output': 'output.h5ad'
 }
 meta = {
-    'functionality_name': 'foo',
+    'name': 'foo',
 }
 ## VIASH END
 
@@ -21,8 +22,20 @@
 classifier.fit(input_train.obsm["X_pca"], input_train.obs["label"].astype(str))
 
 print("Predict on test data", flush=True)
-input_test.obs["label_pred"] = classifier.predict(input_test.obsm["X_pca"])
+label_pred = classifier.predict(input_test.obsm["X_pca"])
 
-print("Write output to file", flush=True)
-input_test.uns["method_id"] = meta["functionality_name"]
-input_test.write_h5ad(par['output'], compression="gzip")
+print("Create output data", flush=True)
+output = ad.AnnData(
+    obs=pd.DataFrame(
+        { 'label_pred': label_pred },
+        index=input_test.obs.index
+    ),
+    uns={
+        'method_id': meta['name'],
+        "dataset_id": input_test.uns["dataset_id"],
+        "normalization_id": input_test.uns["normalization_id"]
+    }
+)
+
+print("Write output data", flush=True)
+output.write_h5ad(par['output'], compression="gzip")

src/methods/logistic_regression/script.py

@@ -1,3 +1,4 @@
+import pandas as pd
 import anndata as ad
 import sklearn.linear_model
 
@@ -8,7 +9,7 @@
     'output': 'output.h5ad'
 }
 meta = {
-    'functionality_name': 'foo',
+    'name': 'foo',
 }
 ## VIASH END
 
@@ -21,8 +22,20 @@
 classifier.fit(input_train.obsm["X_pca"], input_train.obs["label"].astype(str))
 
 print("Predict on test data", flush=True)
-input_test.obs["label_pred"] = classifier.predict(input_test.obsm["X_pca"])
+label_pred = classifier.predict(input_test.obsm["X_pca"])
 
-print("Write output to file", flush=True)
-input_test.uns["method_id"] = meta["functionality_name"]
-input_test.write_h5ad(par['output'], compression="gzip")
+print("Create output data", flush=True)
+output = ad.AnnData(
+    obs=pd.DataFrame(
+        { 'label_pred': label_pred },
+        index=input_test.obs.index
+    ),
+    uns={
+        'method_id': meta['name'],
+        "dataset_id": input_test.uns["dataset_id"],
+        "normalization_id": input_test.uns["normalization_id"]
+    }
+)
+
+print("Write output data", flush=True)
+output.write_h5ad(par['output'], compression="gzip")

src/methods/mlp/script.py

@@ -1,3 +1,4 @@
+import pandas as pd
 import anndata as ad
 from sklearn.neural_network import MLPClassifier
 
@@ -24,8 +25,20 @@
 classifier.fit(input_train.obsm["X_pca"], input_train.obs["label"].astype(str))
 
 print("Predict on test data", flush=True)
-input_test.obs["label_pred"] = classifier.predict(input_test.obsm["X_pca"])
+label_pred = classifier.predict(input_test.obsm["X_pca"])
 
-print("Write output to file", flush=True)
-input_test.uns["method_id"] = meta["functionality_name"]
-input_test.write_h5ad(par['output'], compression="gzip")
+print("Create output data", flush=True)
+output = ad.AnnData(
+    obs=pd.DataFrame(
+        { 'label_pred': label_pred },
+        index=input_test.obs.index
+    ),
+    uns={
+        'method_id': meta['name'],
+        "dataset_id": input_test.uns["dataset_id"],
+        "normalization_id": input_test.uns["normalization_id"]
+    }
+)
+
+print("Write output data", flush=True)
+output.write_h5ad(par['output'], compression="gzip")

src/methods/naive_bayes/script.py

@@ -1,3 +1,4 @@
+import pandas as pd
 import anndata as ad
 import sklearn.naive_bayes
 
@@ -8,7 +9,7 @@
     'output': 'output.h5ad'
 }
 meta = {
-    'functionality_name': 'foo',
+    'name': 'foo',
 }
 ## VIASH END
 
@@ -21,8 +22,20 @@
 classifier.fit(input_train.obsm["X_pca"], input_train.obs["label"].astype(str))
 
 print("Predict on test data", flush=True)
-input_test.obs["label_pred"] = classifier.predict(input_test.obsm["X_pca"])
+label_pred = classifier.predict(input_test.obsm["X_pca"])
 
-print("Write output to file", flush=True)
-input_test.uns["method_id"] = meta["functionality_name"]
-input_test.write_h5ad(par['output'], compression="gzip")
+print("Create output data", flush=True)
+output = ad.AnnData(
+    obs=pd.DataFrame(
+        { 'label_pred': label_pred },
+        index=input_test.obs.index
+    ),
+    uns={
+        'method_id': meta['name'],
+        "dataset_id": input_test.uns["dataset_id"],
+        "normalization_id": input_test.uns["normalization_id"]
+    }
+)
+
+print("Write output data", flush=True)
+output.write_h5ad(par['output'], compression="gzip")

src/methods/scanvi_scarches/script.py

@@ -1,5 +1,5 @@
 import anndata as ad
-import numpy as np
+import pandas as pd
 import scvi
 
 ## VIASH START
@@ -52,10 +52,21 @@
 train_kwargs = dict(max_epochs=par["max_epochs"], early_stopping=True)
 query_model.train(plan_kwargs=dict(weight_decay=0.0), **train_kwargs)
 
-print("Generate predictions", flush=True)
-input_test.obs["label"] = "Unknown"
-input_test.obs["label_pred"] = query_model.predict(input_test)
+print("Predict on test data", flush=True)
+label_pred = query_model.predict(input_test)
 
-print("Write output AnnData to file", flush=True)
-input_test.uns["method_id"] = meta["functionality_name"]
-input_test.write_h5ad(par["output"], compression="gzip")
+print("Create output data", flush=True)
+output = ad.AnnData(
+    obs=pd.DataFrame(
+        { 'label_pred': label_pred },
+        index=input_test.obs.index
+    ),
+    uns={
+        'method_id': meta['name'],
+        "dataset_id": input_test.uns["dataset_id"],
+        "normalization_id": input_test.uns["normalization_id"]
+    }
+)
+
+print("Write output data", flush=True)
+output.write_h5ad(par['output'], compression="gzip")

src/methods/seurat_transferdata/script.R

@@ -10,6 +10,9 @@ par <- list(
   input_test = "resources_test/label_projection/pancreas/test.h5ad",
   output = "output.h5ad"
 )
+meta <- list(
+  name = "seurat_transferdata"
+)
 ## VIASH END
 
 packageVersion("Matrix")
@@ -18,11 +21,6 @@ cat(">> Load input data\n")
 input_train <- read_h5ad(par$input_train)
 input_test <- read_h5ad(par$input_test)
 
-# sce_train <- zellkonverter::readH5AD(par$input_train)
-# obj_train <- Seurat::as.Seurat(sce_train, data = "normalized")
-# sce_test <- zellkonverter::readH5AD(par$input_test)
-# obj_test <- Seurat::as.Seurat(sce_test, data = "normalized")
-
 cat(">> Converting AnnData to Seurat\n")
 anndataToSeurat <- function(adata) {
   # interpreted from https://github.com/satijalab/seurat/blob/v3.1.0/R/objects.R
@@ -31,7 +29,7 @@ anndataToSeurat <- function(adata) {
       counts = as(Matrix::t(adata$layers[["counts"]]), "CsparseMatrix")
     ) %>%
     SeuratObject::SetAssayData(
-      slot = "data",
+      layer = "data",
       new.data = as(Matrix::t(adata$layers[["normalized"]]), "CsparseMatrix")
     ) %>%
     SeuratObject::AddMetaData(
@@ -43,8 +41,10 @@ anndataToSeurat <- function(adata) {
 
   # set embedding
   # could add loadings and stdev
+  X_pca <- adata$obsm[["X_pca"]]
+  dimnames(X_pca) <- list(rownames(adata), paste0("PC_", seq_len(ncol(X_pca))))
   embed <- SeuratObject::CreateDimReducObject(
-    embeddings = adata$obsm[["X_pca"]],
+    embeddings = X_pca,
     key = "PC_"
   )
   obj[["pca"]] <- embed
@@ -74,8 +74,20 @@ query <- Seurat::TransferData(
   refdata = list(labels = "label"),
   verbose = FALSE
 )
-input_test$obs[["label_pred"]] <- query$predicted.labels[input_test$obs_names]
+
+cat(">> Create output data\n")
+output <- anndata::AnnData(
+  obs = data.frame(
+    row.names = input_test$obs_names,
+    label_pred = query$predicted.labels
+  ),
+  uns = list(
+    method_id = meta$name,
+    dataset_id = input_test$uns[["dataset_id"]],
+    normalization_id = input_test$uns[["normalization_id"]]
+  ),
+  shape = c(input_test$n_obs, 0L)
+)
 
 cat(">> Write output to file\n")
-input_test$uns[["method_id"]] <- meta[["functionality_name"]]
-input_test$write_h5ad(par$output, compression = "gzip")
+output$write_h5ad(par$output, compression = "gzip")

src/methods/xgboost/script.py

@@ -1,3 +1,4 @@
+import pandas as pd
 import anndata as ad
 import xgboost as xgb
 
@@ -8,7 +9,7 @@
     'output': 'output.h5ad'
 }
 meta = {
-    'functionality_name': 'foo',
+    'name': 'foo'
 }
 ## VIASH END
 
@@ -32,8 +33,20 @@
 
 print("Predict on test data", flush=True)
 pred = xgb_op.predict(xg_test).astype(int)
-input_test.obs["label_pred"] = categories[pred]
+label_pred = categories[pred]
 
-print("Write output to file", flush=True)
-input_test.uns["method_id"] = meta["functionality_name"]
-input_test.write_h5ad(par['output'], compression="gzip")
+print("Create output data", flush=True)
+output = ad.AnnData(
+    obs=pd.DataFrame(
+        { 'label_pred': label_pred },
+        index=input_test.obs.index
+    ),
+    uns={
+        'method_id': meta['name'],
+        "dataset_id": input_test.uns["dataset_id"],
+        "normalization_id": input_test.uns["normalization_id"]
+    }
+)
+
+print("Write output data", flush=True)
+output.write_h5ad(par['output'], compression="gzip")