Removed umap correlation

mehta-lab · Sep 25, 2024 · 3e215b5 · 3e215b5
1 parent a9a4ab8
commit 3e215b5
Show file tree

Hide file tree

Showing 2 changed files with 6 additions and 81 deletions.
diff --git a/applications/contrastive_phenotyping/evaluation/PC_vs_CF.py b/applications/contrastive_phenotyping/evaluation/PC_vs_CF.py
@@ -6,6 +6,7 @@
 # %%
 from pathlib import Path
 import sys
+import os
 
 sys.path.append("/hpc/mydata/soorya.pradeep/scratch/viscy_infection_phenotyping/VisCy")
 
@@ -50,19 +51,8 @@
 embedding_dataset = read_embedding_dataset(features_path)
 embedding_dataset
 
-# %% umap analsis of the features
-
 # load all unprojected features:
 features = embedding_dataset["features"]
-scaled_features = StandardScaler().fit_transform(features.values)
-
-umap = UMAP()
-embedding = umap.fit_transform(features.values)
-features = (
-    features.assign_coords(UMAP1=("sample", embedding[:, 0]))
-    .assign_coords(UMAP2=("sample", embedding[:, 1]))
-    .set_index(sample=["UMAP1", "UMAP2"], append=True)
-)
 
 # %% PCA analysis of the features
 
@@ -325,14 +315,14 @@
 # %% plot PCA vs set of computed features
 
 set_features = [
-    "Sensor Area",
     "Fluor radial profile",
-    "Contrast Fluor",
-    "Phase radial profile",
-    "Phase Standard Deviation",
-    "Phase IQR",
     "Homogeneity Phase",
+    "Phase IQR",
+    "Phase Standard Deviation",
+    "Sensor Area",
     "Homogeneity Fluor",
+    "Contrast Fluor",
+    "Phase radial profile",
 ]
 
 plt.figure(figsize=(8, 10))
@@ -349,27 +339,8 @@
     "/hpc/projects/comp.micro/infected_cell_imaging/Single_cell_phenotyping/ContrastiveLearning/Figure_panels/cell_division/PC_vs_CF_2chan_pca_setfeatures.svg"
 )
 
-# %% display UMAP correlation as a heatmap
-
-plt.figure(figsize=(20, 5))
-sns.heatmap(
-    correlation.drop(columns=["UMAP1", "UMAP2"]).loc["UMAP1":"UMAP2", :],
-    annot=True,
-    cmap="coolwarm",
-    fmt=".2f",
-)
-plt.title("Correlation between UMAP features and computed features")
-plt.xlabel("Computed Features")
-plt.ylabel("UMAP Features")
-plt.savefig(
-    "/hpc/projects/comp.micro/infected_cell_imaging/Single_cell_phenotyping/ContrastiveLearning/Figure_panels/cell_division/PC_vs_CF_2chan.svg"
-)
-
 # %% find the cell patches with the highest and lowest value in each feature
 
-import os
-
-
 def save_patches(fov_name, track_id):
     data_path = Path(
         "/hpc/projects/intracellular_dashboard/viral-sensor/2024_02_04_A549_DENV_ZIKV_timelapse/8-train-test-split/registered_test.zarr"

diff --git a/applications/contrastive_phenotyping/evaluation/PC_vs_CF_singleChannel.py b/applications/contrastive_phenotyping/evaluation/PC_vs_CF_singleChannel.py
@@ -50,19 +50,8 @@
 embedding_dataset = read_embedding_dataset(features_path)
 embedding_dataset
 
-# %% umap analsis of the features
-
 # load all unprojected features:
 features = embedding_dataset["features"]
-scaled_features = StandardScaler().fit_transform(features.values)
-
-umap = UMAP()
-embedding = umap.fit_transform(features.values)
-features = (
-    features.assign_coords(UMAP1=("sample", embedding[:, 0]))
-    .assign_coords(UMAP2=("sample", embedding[:, 1]))
-    .set_index(sample=["UMAP1", "UMAP2"], append=True)
-)
 
 # %% PCA analysis of the features
 
@@ -214,41 +203,6 @@
 # Compute correlation between PCA features and computed features
 correlation = feature_df_removed.corr(method="spearman")
 
-# %% display PCA correlation as a heatmap
-
-# remove UMAP compnents from the correlation matrix before plotting
-correlation_pca = correlation.drop(columns=["UMAP1", "UMAP2"])
-
-plt.figure(figsize=(20, 5))
-sns.heatmap(
-    correlation_pca.drop(columns=["PCA1", "PCA2", "PCA3"]).loc["PCA1":"PCA3", :],
-    annot=True,
-    cmap="coolwarm",
-    fmt=".2f",
-)
-plt.title("Correlation between PCA features and computed features")
-plt.xlabel("Computed Features")
-plt.ylabel("PCA Features")
-plt.savefig(
-    "/hpc/projects/comp.micro/infected_cell_imaging/Single_cell_phenotyping/ContrastiveLearning/Figure_panels/cell_division/PC_vs_CF_umap_phaseonly.svg"
-)
-
-# %% display UMAP correlation as a heatmap
-
-plt.figure(figsize=(20, 5))
-sns.heatmap(
-    correlation.drop(columns=["UMAP1", "UMAP2"]).loc["UMAP1":"UMAP2", :],
-    annot=True,
-    cmap="coolwarm",
-    fmt=".2f",
-)
-plt.title("Correlation between UMAP features and computed features")
-plt.xlabel("Computed Features")
-plt.ylabel("UMAP Features")
-plt.savefig(
-    "/hpc/projects/comp.micro/infected_cell_imaging/Single_cell_phenotyping/ContrastiveLearning/Figure_panels/cell_division/PC_vs_CF_phaseonly.svg"
-)
-
 # %% calculate the p-value and draw volcano plot to show the significance of the correlation
 
 p_values = pd.DataFrame(index=correlation.index, columns=correlation.columns)