extra-trees-regressor model for evaluation

et/cl_features.py

@@ -0,0 +1,60 @@
+import coderdata as cd
+import pandas as pd
+import matplotlib.pyplot as plt
+import numpy as np
+import umap
+import seaborn as sns
+import matplotlib.patches as mpatches
+import numpy as np
+import pickle
+import matplotlib.pyplot as plt
+from sklearn.preprocessing import StandardScaler
+# from sklearn.ensemble import ExtraTreesClassifier
+from sklearn import preprocessing
+
+
+hcmi = cd.DatasetLoader('hcmi')
+beataml = cd.DatasetLoader('beataml')
+cptac = cd.DatasetLoader('cptac')
+depmap = cd.DatasetLoader('broad_sanger')
+mpnst = cd.DatasetLoader('mpnst')
+
+# Join BeatAML and HCMI
+joined_dataset0 = cd.join_datasets(beataml, hcmi)
+# Join DepMap and CPTAC
+joined_dataset1 = cd.join_datasets(depmap, cptac)
+# Join Datasets
+joined_dataset2 = cd.join_datasets(joined_dataset0,joined_dataset1)
+# Final Join
+joined_dataset3 = cd.join_datasets(joined_dataset2,mpnst)
+joined_dataset3.transcriptomics= joined_dataset3.transcriptomics[["improve_sample_id", "transcriptomics", "entrez_id", "source", "study"]]
+
+
+# select target type
+data = joined_dataset3.experiments[joined_dataset3.experiments.dose_response_metric == 'auc']
+# select study
+ctrpv2 = data[data.study == 'CTRPv2']
+# joined_dataset3.transcriptomics.study.unique()
+
+# get cancer type map
+cancer_type_sample_map = dict(zip(joined_dataset3.samples['improve_sample_id'], joined_dataset3.samples['cancer_type']))
+
+# pivot
+cl = joined_dataset3.transcriptomics.pivot_table(index='improve_sample_id', columns='entrez_id', values='transcriptomics')
+cl = cl.dropna(axis=1)
+cl.columns.name = None
+cl.reset_index(inplace=True)
+
+
+
+# add cancer type
+cl['cancer_type'] = cl.improve_sample_id.map(cancer_type_sample_map)
+
+# convert str labels to int
+le = preprocessing.LabelEncoder()
+le.fit(cl.cancer_type)
+cl['cancer_type_int'] = le.transform(cl.cancer_type)
+cl.to_csv('cl_features/cl.csv', index=False)
+
+
+

et/et.py

@@ -0,0 +1,108 @@
+import coderdata as cd
+import pandas as pd
+import matplotlib.pyplot as plt
+import numpy as np
+import umap
+import seaborn as sns
+import matplotlib.patches as mpatches
+import numpy as np
+import pickle
+import matplotlib.pyplot as plt
+from sklearn.preprocessing import StandardScaler
+from sklearn.ensemble import ExtraTreesClassifier
+from sklearn import preprocessing
+from sklearn.metrics import r2_score, mean_squared_error
+from sklearn.ensemble import ExtraTreesRegressor
+from sklearn.preprocessing import StandardScaler
+import numpy as np
+from sklearn.model_selection import train_test_split
+from sklearn.ensemble import ExtraTreesRegressor
+
+
+hcmi = cd.DatasetLoader('hcmi')
+beataml = cd.DatasetLoader('beataml')
+cptac = cd.DatasetLoader('cptac')
+depmap = cd.DatasetLoader('broad_sanger')
+mpnst = cd.DatasetLoader('mpnst')
+
+# Join BeatAML and HCMI
+joined_dataset0 = cd.join_datasets(beataml, hcmi)
+# Join DepMap and CPTAC
+joined_dataset1 = cd.join_datasets(depmap, cptac)
+# Join Datasets
+joined_dataset2 = cd.join_datasets(joined_dataset0,joined_dataset1)
+# Final Join
+joined_dataset3 = cd.join_datasets(joined_dataset2,mpnst)
+joined_dataset3.transcriptomics= joined_dataset3.transcriptomics[["improve_sample_id", "transcriptomics", "entrez_id", "source", "study"]]
+
+data = joined_dataset3.experiments[joined_dataset3.experiments.dose_response_metric == 'auc']
+data.improve_drug_id.nunique()
+
+ctrpv2 = data[data.study == 'CTRPv2']
+ctrpv2.to_csv('cl_features/data.csv', index=False)
+ctrpv2 = pd.read_csv('cl_features/data.csv')
+
+
+
+#### load cl data
+
+cl = pd.read_csv('cl_features/cl.csv')
+cl_features = pd.read_pickle('cl_features/cl_1000.pkl')
+# list(cl_features)[0]
+cl_features = [str(i) for i in list(cl_features)]
+cl = cl.loc[:, ['improve_sample_id'] + cl_features]
+
+
+
+md = pd.read_csv('drugs/mdm.csv')
+v = md.iloc[:, :-1]
+remove=[]
+for i in v.columns:
+
+    v2 = v.loc[:, i]
+    try:
+        [float(i) for i in v2.values]
+    except:
+        remove.append(i)
+
+
+v = md.drop(remove, axis=1)
+
+id2smiles = dict(joined_dataset3.drugs[['improve_drug_id', 'canSMILES']].drop_duplicates().values)
+smiles2id = {v:k for k,v in id2smiles.items()}
+v['improve_drug_id'] = [smiles2id[i] for i in v.smiles]
+
+v2 = v.drop(columns=['smiles'], axis=1)
+
+# np.where(np.isnan(v3))
+# sc = StandardScaler()
+
+
+data2 = ctrpv2[['improve_sample_id', 'improve_drug_id','dose_response_value']]
+d2 = pd.merge(data2, cl, on='improve_sample_id', how='left')
+d3 = pd.merge(d2, v2, on='improve_drug_id', how='left')
+d3 = d3.dropna(axis=0)
+
+
+# sc = StandardScaler()
+d3.reset_index(drop=True, inplace=True)
+train, test = train_test_split(d3, test_size=.1)
+# test, val = train_test_split(val, test_size=.5)
+
+train_tmp = train.sample(frac=.1)
+# train_x = train.iloc[:, 3:]
+train_x = train_tmp.iloc[:, 3:]
+test_x = test.iloc[:, 3:]
+
+# train_y = train.dose_response_value.values
+train_y = train_tmp.dose_response_value.values
+test_y = test.dose_response_value.values
+
+sc = StandardScaler()
+train_x = sc.fit_transform(train_x)
+test_x = sc.transform(test_x)
+
+
+et = ExtraTreesRegressor()
+et.fit(train_x, train_y)
+et.predict(test_x)