Merge pull request #244 from DendrouLab/fc_scviseeds

adding seeds to each scvi-tool script

CHANGELOG.md

@@ -17,6 +17,7 @@
 ### dependencies
 - All the dependencies have been updated. 
   - Python>=3.10 required
+- added seeds to all scvi tasks for reproducibility
 
 ## v0.4.1
 

docs/release_notes.md

@@ -17,6 +17,7 @@
 ### dependencies
 - All the dependencies have been updated. 
   - Python>=3.10 required
+- added seeds to all scvi tasks for reproducibility
 
 ## v0.4.1
 

docs/yaml_docs/useful_info_on_yml.md

@@ -83,12 +83,23 @@ Therefore, `PARAMS['prot']['neighbors']` is the equivalent of `prot_params['neig
 
 You can read the `pipeline.yml` configuration file in python and check how the commands are parsed using
 
+```python
+import yaml
+with open('pipeline.yml', 'r') as file:
+    PARAMS = yaml.safe_load(file)
+
 ```
-from panpipes.funcs.io import read_yaml
 
-PARAMS = read_yaml("pipeline.yml")
+in `panpipes` we read the yaml configuration file at the beginning of each workflow with:
+
+```python
+from cgatcore import pipeline as P
+PARAMS = P.get_parameters(
+    ["%s/pipeline.yml" % os.path.splitext(__file__)[0],
+     "pipeline.yml"])
 ```
 
+
 ### Anchors and Scalars
 
 YAML files offer a lot of functionalities to fill and parse blocks of text avoiding to repeat entire sections. 

panpipes/panpipes/pipeline_integration.py

@@ -228,6 +228,9 @@ def run_scvi(outfile):
      --figdir figures/rna/
      """
     # cannot use the normal method for importing params from yaml, because it only works up to depth 2
+    # fetch seeds
+    scvi_params =  PARAMS['rna']['scvi']
+    cmd += " --scvi_seed %i" % scvi_params['seed']
     neighbor_params = PARAMS['rna']['neighbors']
     if neighbor_params['method'] is not None:
         cmd += " --neighbors_method %s" % neighbor_params['method']
@@ -550,8 +553,12 @@ def run_totalvi(outfile):
      --output_csv %(outfile)s 
      --figdir figures/
      """
+    scvi_params =  PARAMS['multimodal']['totalvi']
+    cmd += " --scvi_seed %i" % scvi_params['seed']
+
+
     if PARAMS['multimodal_column_categorical'] is not None:
-        cmd += "--integration_col_categorical %(multimodal_column_categorical)s "
+        cmd += " --integration_col_categorical %(multimodal_column_categorical)s "
     neighbor_params = PARAMS['multimodal']['neighbors']
     if neighbor_params['method'] is not None:
         cmd += " --neighbors_method %s" % neighbor_params['method']
@@ -587,9 +594,12 @@ def run_multivi(outfile):
      --output_csv %(outfile)s 
      --figdir figures/
      """
+    scvi_params =  PARAMS['multimodal']['MultiVI']
+    cmd += " --scvi_seed %i" % scvi_params['seed']
 
+
     if PARAMS['multimodal_column_categorical'] is not None:
-        cmd += "--integration_col_categorical %(multimodal_column_categorical)s "
+        cmd += " --integration_col_categorical %(multimodal_column_categorical)s "
 
     if PARAMS['multimodal_column_continuous'] is not None:
         cmd += "--integration_col_continuous %(multimodal_column_continuous)s "

panpipes/panpipes/pipeline_integration/pipeline.yml

@@ -26,15 +26,16 @@ queues:
 # --------------------------------
 
 # ----------------------------
-# Data format
+# Data input
+# ----------------------------
 sample_prefix: test
 preprocessed_obj: ../preprocess/test.h5mu
 
 
 #-----------------
 # Batch correction
 # ----------------
-# Batch correction is done unimodal, meaning each modality is batch corrected independently
+# Batch correction, unimodal correction: meaning each modality is batch corrected independently.
 
 # ------------
 # RNA modality
@@ -55,6 +56,7 @@ rna:
 
   # SCVI args
   scvi:
+    seed: 1492
     exclude_mt_genes: True
     mt_column: mt
     model_args:
@@ -144,6 +146,7 @@ multimodal:
 
   # TotalVI arguments
   totalvi:
+    seed: 1492
     modalities: rna,prot
     exclude_mt_genes: True
     mt_column: mt
@@ -159,6 +162,7 @@ multimodal:
 
   # MultiVI arguments
   MultiVI:
+    seed: 1492
     lowmem: True
     model_args:
       n_hidden :

panpipes/python_scripts/batch_correct_multivi.py

@@ -46,6 +46,8 @@
                     help="neighbors k")
 parser.add_argument('--neighbors_metric',default="euclidean",
                     help="neighbor metric, e.g. euclidean or cosine")
+parser.add_argument('--scvi_seed',default=None,
+                    help="set explicitly seed to make runs reproducible")
 
 
 
@@ -57,7 +59,10 @@
 sc.settings.autoshow = False
 sc.settings.figdir = args.figdir
 
-scvi.settings.seed = 1492
+if args.scvi_seed is not None:
+    scvi.settings.seed = int(args.scvi_seed)
+else:
+    scvi.settings.seed = 1492
 # load parameters
 
 threads_available = multiprocessing.cpu_count()

panpipes/python_scripts/batch_correct_scvi.py

@@ -44,6 +44,9 @@
                     help="neighbors k")
 parser.add_argument('--neighbors_metric',
                     help="neighbor metric, e.g. euclidean or cosine")
+parser.add_argument('--scvi_seed',default=None,
+                    help="set explicitly seed to make runs reproducible")
+
 
 
 args, opt = parser.parse_known_args()
@@ -61,6 +64,12 @@
 sc.settings.autoshow = False
 sc.settings.figdir = args.figdir
 
+if args.scvi_seed is not None:
+    scvi.settings.seed = int(args.scvi_seed)
+else:
+    scvi.settings.seed = 1492
+
+
 
 # test_script=params['rna']['scvi']['testrun']
 test_script=False

panpipes/python_scripts/batch_correct_totalvi.py

@@ -46,6 +46,11 @@
                     help="neighbors k")
 parser.add_argument('--neighbors_metric',default="euclidean",
                     help="neighbor metric, e.g. euclidean or cosine")
+parser.add_argument('--scvi_seed',default=None,
+                    help="set explicitly seed to make runs reproducible")
+
+
+
 
 args, opt = parser.parse_known_args()
 L.info(args)
@@ -55,6 +60,11 @@
 sc.settings.figdir = args.figdir
 
 
+if args.scvi_seed is not None:
+    scvi.settings.seed = int(args.scvi_seed)
+else:
+    scvi.settings.seed = 1492
+
 # load parameters
 
 threads_available = multiprocessing.cpu_count()

panpipes/python_scripts/plot_qc_spatial.py

@@ -52,8 +52,8 @@
 
 figdir = args.figdir
 
-if not os.path.exists(figdir):
-    os.mkdir(figdir)
+# if not os.path.exists(figdir):
+#     os.mkdir(figdir)
 
 sc.settings.figdir = figdir
 sc.set_figure_params(scanpy=True, fontsize=14, dpi=300, facecolor='white', figsize=(5,5))

tests/integration_1/pipeline.yml

@@ -215,6 +215,7 @@ multimodal:
   column_categorical: dataset
   # extra params:
   totalvi:
+    seed: 1492
     # this is a minimal set of parameters that will be expected
     # you can add any other param from the tutorials and they will
     # be parsed alongside the others
@@ -234,6 +235,7 @@ multimodal:
       early_stopping: True
     training_plan: None
   MultiVI:
+    seed: 1492
     # this is a minimal set of parameters that will be expected
     # you can add any other param from the tutorials and they will
     # be parsed alongside the others
@@ -375,6 +377,7 @@ plotqc:
 # Choose the integration results you want to merge in the final object
 # For unimodal integration: to pick the uncorrected version use "no_correction"
 # then run
+
 # panpipes integration make merge_integration
 final_obj:
   rna:

tests/integration_2/pipeline.yml

@@ -215,6 +215,7 @@ multimodal:
   column_categorical: dataset
   # extra params:
   totalvi:
+    seed: 1492
     # this is a minimal set of parameters that will be expected
     # you can add any other param from the tutorials and they will
     # be parsed alongside the others
@@ -234,6 +235,7 @@ multimodal:
       early_stopping: True
     training_plan: None
   MultiVI:
+    seed: 1492
     # this is a minimal set of parameters that will be expected
     # you can add any other param from the tutorials and they will
     # be parsed alongside the others