def adam_pretrain(model, model_name, train_xs, train_ys, num_epoch, test_xs, test_ys):
model.compile(optimizer=keras.optimizers.Adam(),
loss='categorical_crossentropy', metrics=['accuracy'])
model.fit(train_xs, train_ys, batch_size=128, nb_epoch=num_epoch,
validation_data=(test_xs, test_ys), shuffle=True)
model_name = '%s_adam_pretrain' % model_name
save_model(model, model_name)
model = load_model(model_name)
return model
n_iter = 300
if classification_task:
rf_gs = supervised_learning_steps("rf", "roc_auc", data_type,
classification_task, model, param_rf,
X_train, y_train, n_iter)
else:
rf_gs = supervised_learning_steps("rf", "r2", data_type,
classification_task, model, param_rf,
X_train, y_train, n_iter)
rf_gs.cv_results_
# +
rf_gs = load_model("rf_models/rf_" + data_type_options[input_option] +
"_regressor_gs.pk")
np.max(rf_gs.cv_results_["mean_test_score"])
file_list = [
"../data/Test_Compound_Viral_interactions_for_Supervised_Learning_with_LS_LS.csv",
"../data/Test_Compound_Viral_interactions_for_Supervised_Learning_with_MFP_LS.csv"
]
filename = file_list[input_option]
with open(filename, "rb") as file:
print("Loading ", filename)
big_df = pd.read_csv(filename, header='infer', delimiter=",")
total_length = len(big_df.columns)
X = big_df.iloc[:, range(5, total_length)]
Y = big_df[['pchembl_value']].to_numpy().flatten()
meta_X = big_df.iloc[:, [0, 1, 2, 3]]
# -
def get_cv_results(model, index, cv_scores, nfolds):
for i in range(0, nfolds):
cv_scores.append(
np.round(
model.cv_results_["split" + str(i) + "_test_score"][index],
nfolds))
return (cv_scores)
# +
#Get the CV models and optimal scores
#GLM with protein and compound LS
glm_LS_LS = load_model(
"../models/glm_models/glm_LS_Compound_LS_Protein_regressor_gs.pk")
glm_LS_LS_index = np.argmax(glm_LS_LS.cv_results_["mean_test_score"])
glm_LS_LS_r2_scores = get_cv_results(glm_LS_LS, glm_LS_LS_index, [], 5)
glm_MFP_LS = load_model(
"../models/glm_models/glm_MFP_Compound_LS_Protein_regressor_gs.pk")
glm_MFP_LS_index = np.argmax(glm_MFP_LS.cv_results_["mean_test_score"])
glm_MFP_LS_r2_scores = get_cv_results(glm_MFP_LS, glm_MFP_LS_index, [], 5)
#RF with protein and compound LS
rf_LS_LS = load_model(
"../models/rf_models/rf_LS_Compound_LS_Protein_regressor_gs.pk")
rf_LS_LS_index = np.argmax(rf_LS_LS.cv_results_["mean_test_score"])
rf_LS_LS_r2_scores = get_cv_results(rf_LS_LS, rf_LS_LS_index, [], 5)
rf_MFP_LS = load_model(
X_train_copy = scaler.fit_transform(X_train)
if classification_task:
svm_gs = supervised_learning_steps("svm", "roc_auc", data_type,
classification_task, model, param_svm,
X_train_copy, y_train, n_iter)
else:
svm_gs = supervised_learning_steps("svm", "r2", data_type,
classification_task, model, param_svm,
X_train_copy, y_train, n_iter)
svm_gs.cv_results_
save_model(scaler, "%s_models/%s_%s_scaling_gs.pk" % ("svm", "svm", data_type))
# -
svm_gs = load_model("svm_models/svm__LS_Drug_LS_Protein_regressor_gs.pk")
scaler = load_model("svm_models/svm__LS_Drug_LS_Protein_scaling_gs.pk")
svm_best = svm_gs.best_estimator_
# +
np.max(svm_gs.cv_results_['mean_test_score'])
file_list = [
"../data/Test_Compound_Viral_interactions_for_Supervised_Learning_with_LS_LS.csv",
"../data/Test_Compound_Viral_interactions_for_Supervised_Learning_with_MFP_LS.csv"
]
filename = file_list[input_option]
with open(filename, "rb") as file:
print("Loading ", filename)
big_df = pd.read_csv(filename, header='infer', delimiter=",")
print(X_train[:10])
print(X_train.shape, y_train.shape)
print(X_train.columns)
print(big_df.isnull().sum().sum())
print("Loaded training file")
#Get results for test file
print("Loading test file")
test_filename = args.input2
big_X_test = pd.read_csv("../data/" + args.input2, header='infer', sep=",")
total_length = len(big_X_test.columns)
X_test = big_X_test.iloc[:, range(5, total_length)]
if ("MFP" not in args.input1):
rf_best = load_model(
"../models/rf_models/rf_LS_Compound_LS_Protein_regressor_best_estimator.pk"
)
else:
rf_best = load_model(
"../models/rf_models/rf_MFP_Compound_LS_Protein_regressor_best_estimator.pk"
)
print("Making Predictions")
y_pred = rf_best.predict(X_test)
meta_X_test = big_X_test.iloc[:, [0, 2]].copy()
meta_X_test.loc[:, 'predictions'] = y_pred
if ("sars_cov_2" in args.input2):
meta_X_test.loc[:, 'labels'] = 0
else:
meta_X_test.loc[:, 'labels'] = big_X_test.iloc[:, 4].copy()
print(X_train[:10])
print(X_train.shape,y_train.shape)
print(X_train.columns)
print(big_df.isnull().sum().sum())
print("Loaded training file")
#Get results for test file
print("Loading test file")
test_filename = args.input2
big_X_test = pd.read_csv("../data/"+args.input2,header='infer',sep=",")
total_length = len(big_X_test.columns)
X_test = big_X_test.iloc[:,range(5,total_length)]
if ("MFP" not in args.input1):
xgb_best = load_model("../models/xgb_models/xgb_LS_Compound_LS_Protein_regressor_best_estimator.pk")
else:
xgb_best = load_model("../models/xgb_models/xgb_MFP_Compound_LS_Protein_regressor_best_estimator.pk")
print("Making Predictions")
y_pred = xgb_best.predict(X_test)
meta_X_test = big_X_test.iloc[:,[0,2]].copy()
meta_X_test.loc[:,'predictions']=y_pred
if ("sars_cov_2" in args.input2):
meta_X_test.loc[:,'labels']=0
else:
meta_X_test.loc[:,'labels']=big_X_test.iloc[:,4].copy()
out_file = args.output
meta_X_test.to_csv("../results/"+out_file,index=False)
## explain the model's predictions using SHAP values
#explainer = shap.TreeExplainer(xgb_gs.best_estimator_)
#shap_values = explainer.shap_values(X_train)
#shap.summary_plot(shap_values, X_train)
# +
#Get results for SARS-COV-2 for SMILES embeddig + protein embedding (input option = 0) or Morgan fingerprints + protein emedding (input_option = 1)
input_option = 0
if (input_option == 0):
big_X_test = pd.read_csv(
"../data/sars_cov_2_Compound_Viral_interactions_for_Supervised_Learning_with_LS_LS.csv",
header='infer',
sep=",")
total_length = len(big_X_test.columns)
X_test = big_X_test.iloc[:, range(5, total_length)]
svm_best = load_model(
"../models/svm_models/svm_LS_Compound_LS_Protein_regressor_best_estimator.pk"
)
scaler = load_model(
"../models/svm_models/svm_LS_Compound_LS_Protein_scaling_gs.pk")
y_pred = svm_best.predict(scaler.transform(X_test))
meta_X_test = big_X_test.iloc[:, [0, 2]].copy()
meta_X_test.loc[:, 'predictions'] = y_pred
meta_X_test.loc[:, 'labels'] = 0
meta_X_test.to_csv("../results/SVM_" + data_type_options[input_option] +
"supervised_sars_cov_2_predictions.csv",
index=False)
elif (input_option == 1):
big_X_test = pd.read_csv(
"../data/sars_cov_2_Compound_Viral_interactions_for_Supervised_Learning_with_MFP_LS.csv",
from keras.datasets import mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()
x = x_train.reshape(-1, 28 * 28)
x = (x-x.mean(axis=1).reshape(-1, 1))/x.std(axis=1).reshape(-1, 1)
x = x.reshape(-1, 28, 28, 1)
y = pd.get_dummies(y_train).to_numpy()
xt = x_test.reshape(-1, 28 * 28)
xt = (xt-xt.mean(axis=1).reshape(-1, 1))/xt.std(axis=1).reshape(-1, 1)
xt = xt.reshape(-1, 28, 28, 1)
yt = pd.get_dummies(y_test).to_numpy()
m = Sequential()
m.add(Conv2d(
input_shape=(28, 28, 1), filters=4, padding=None,
kernel_size=(3, 3), activation="relu"))
m.add(Conv2d(filters=8, kernel_size=(3, 3), padding=None, activation="relu"))
m.add(Pool2d(kernel_size=(2, 2)))
m.add(Flatten())
m.add(FFL(neurons=64, activation="relu"))
m.add(Dropout(0.1))
m.add(FFL(neurons=10, activation='softmax'))
m.compile_model(lr=0.01, opt="adam", loss="cse")
m.summary()
m.train(x[:30], y[:30], epochs=2, batch_size=30, val_x=xt[:10], val_y=yt[:10])
m.visualize()
m.save_model()
load_model()
m.summary()
print(m.predict(x[10]))
X_train_copy = scaler.fit_transform(X_train)
if classification_task:
svm_gs = supervised_learning_steps("svm", "roc_auc", data_type,
classification_task, model, param_svm,
X_train_copy, y_train, n_iter)
else:
svm_gs = supervised_learning_steps("svm", "r2", data_type,
classification_task, model, param_svm,
X_train_copy, y_train, n_iter)
svm_gs.cv_results_
save_model(scaler, "%s_models/%s_%s_scaling_gs.pk" % ("svm", "svm", data_type))
# -
svm_gs = load_model("svm_models/svm__LS_Drug_LS_Protein_regressor_gs.pk")
scaler = load_model("svm_models/svm__LS_Drug_LS_Protein_scaling_gs.pk")
svm_best = svm_gs.best_estimator_
y_pred_svm = svm_best.predict(X_train_copy)
plt.hist(y_pred_svm)
# +
np.max(svm_gs.cv_results_['mean_test_score'])
filename = "../data/Test_Drug_Viral_interactions_with_LS_v2_for_Supervised_Learning.csv"
with open(filename, "rb") as file:
print("Loading ", filename)
big_df = pd.read_csv(filename, header='infer', delimiter=",")
total_length = len(big_df.columns)
X = big_df.iloc[:, range(5, total_length)]
Y = big_df[['pchembl_value']].to_numpy().flatten()
meta_X = big_df.iloc[:, [0, 1, 2, 3]]
n_iter = 300
if classification_task:
rf_gs = supervised_learning_steps("rf", "roc_auc", data_type,
classification_task, model, param_rf,
X_train, y_train, n_iter)
else:
rf_gs = supervised_learning_steps("rf", "r2", data_type,
classification_task, model, param_rf,
X_train, y_train, n_iter)
rf_gs.cv_results_
# +
rf_gs = load_model("rf_models/rf_" + data_type_options[input_option] +
"_regressor_gs.pk")
np.max(rf_gs.cv_results_["mean_test_score"])
file_list = [
"../data/Test_Compound_Viral_interactions_for_Supervised_Learning_with_LS_LS.csv",
"../data/Test_Compound_Viral_interactions_for_Supervised_Learning_with_MFP_LS.csv"
]
filename = file_list[input_option]
with open(filename, "rb") as file:
print("Loading ", filename)
big_df = pd.read_csv(filename, header='infer', delimiter=",")
total_length = len(big_df.columns)
X = big_df.iloc[:, range(5, total_length)]
Y = big_df[['pchembl_value']].to_numpy().flatten()
meta_X = big_df.iloc[:, [0, 1, 2, 3]]
"max_depth": scipy.stats.randint(1, 9),
"min_samples_leaf": scipy.stats.randint(1, 10),
"max_features": scipy.stats.uniform.ppf([0.1,0.7])
}
n_iter=200
if classification_task:
rf_gs=supervised_learning_steps("rf","roc_auc",data_type,classification_task,model,param_rf,X_train,y_train,n_iter)
else:
rf_gs=supervised_learning_steps("rf","r2",data_type,classification_task,model,param_rf,X_train,y_train,n_iter)
rf_gs.cv_results_
# +
rf_gs = load_model("rf_models/rf__LS_Drug_LS_Protein_regressor_gs.pk")
np.max(rf_gs.cv_results_["mean_test_score"])
filename = "../data/Test_Drug_Viral_interactions_with_LS_v2_for_Supervised_Learning.csv"
with open(filename, "rb") as file:
print("Loading ", filename)
big_df = pd.read_csv(filename, header='infer', delimiter=",")
total_length = len(big_df.columns)
X = big_df.iloc[:,range(5,total_length)]
Y = big_df[['pchembl_value']].to_numpy().flatten()
meta_X = big_df.iloc[:,[0,1,2,3]]
print("Lengths --> X = %d, Y = %d" % (len(X), len(Y)))
print(X.columns)
n_samples = len(X)
indices = np.arange(n_samples)
n_iter = 200
if classification_task:
xgb_gs = supervised_learning_steps("xgb", "roc_auc", data_type,
classification_task, model, param_xgb,
X_train, y_train, n_iter)
else:
xgb_gs = supervised_learning_steps("xgb", "r2", data_type,
classification_task, model, param_xgb,
X_train, y_train, n_iter)
xgb_gs.cv_results_
# -
xgb_gs = load_model("xgb_models/xgb__LS_Drug_LS_Protein_regressor_gs.pk")
xgb_best = xgb_gs.best_estimator_
y_pred_xgb = xgb_best.predict(X_train)
plt.hist(y_pred_xgb)
calculate_regression_metrics(y_train, y_pred_xgb)
# +
np.max(xgb_gs.cv_results_["mean_test_score"])
filename = "../data/Test_Drug_Viral_interactions_with_LS_v2_for_Supervised_Learning.csv"
with open(filename, "rb") as file:
print("Loading ", filename)
big_df = pd.read_csv(filename, header='infer', delimiter=",")
total_length = len(big_df.columns)
X = big_df.iloc[:, range(5, total_length)]
Y = big_df[['pchembl_value']].to_numpy().flatten()
meta_X = big_df.iloc[:, [0, 1, 2, 3]]