Add files via upload

tests/test_regvelo.py

@@ -0,0 +1,39 @@
+## test RegVelo
+import scvelo as scv
+import numpy as np
+import pandas as pd
+from scvi.data import synthetic_iid
+from regvelovi import REGVELOVI,TFscreening
+import torch
+
+def test_regvelo():
+    adata = synthetic_iid()
+    adata.layers["spliced"] = adata.X.copy()
+    adata.layers["unspliced"] = adata.X.copy()
+    adata.var_names = "Gene" + adata.var_names
+    n_gene = len(adata.var_names)
+    ## create W
+    grn_matrix = np.random.choice([0, 1], size=(n_gene,n_gene), p=[0.8, 0.2]).T
+    W = pd.DataFrame(grn_matrix, index=adata.var_names, columns=adata.var_names)
+    adata.uns["skeleton"] = W
+    TF_list = adata.var_names.tolist()
+
+    ## training process
+    W = adata.uns["skeleton"].copy()
+    W = torch.tensor(np.array(W))
+    REGVELOVI.setup_anndata(adata, spliced_layer="spliced", unspliced_layer="unspliced")
+
+    ## Training the model
+    # hard constraint
+    reg_vae = REGVELOVI(adata,W=W.T,regulators = TF_list,soft_constraint = False)
+    reg_vae.train()
+    # soft constraint
+    reg_vae = REGVELOVI(adata,W=W.T,regulators = TF_list,soft_constraint = True)
+    reg_vae.train()
+
+    reg_vae.get_latent_representation()
+    reg_vae.get_velocity()
+    reg_vae.get_latent_time()
+
+    reg_vae.history
+    print(reg_vae)