Disease-gene association code

CODE:

import pandas as pd import numpy as np from sklearn.model_selection import train_test_split, KFold, GridSearchCV from sklearn.preprocessing import StandardScaler from sklearn.svm import SVC from sklearn.ensemble import RandomForestClassifier, GradientBoostingClassifier from sklearn.metrics import accuracy_score, classification_report, make_scorer, f1_score from sklearn.pipeline import Pipeline from sklearn.impute import SimpleImputer import warnings warnings.filterwarnings("ignore") # Raw data string containing disease information data = """ DOID:11054 Urinary bladder cancer 5 of 38 2.13 2.53 1.15e-06 DOID:5672 Large intestine cancer 5 of 44 2.07 2.49 1.15e-06 DOID:5093 Thoracic cancer 5 of 71 1.86 2.13 4.77e-06 DOID:9256 Colorectal cancer 4 of 32 2.11 2.1 1.36e-05 DOID:707 B-cell lymphoma 4 of 34 2.09 2.07 1.57e-05 DOID:9561 Nasopharyngeal disease 3 of 9 2.54 1.99 4.49e-05 DOID:0060119 Pharynx cancer 3 of 12 2.41 1.84 8.75e-05 DOID:0060058 Lymphoma 5 of 105 1.69 1.82 1.77e-05 DOID:3910 Lung adenocarcinoma 3 of 13 2.38 1.82 9.71e-05 DOID:4905 Pancreatic carcinoma 3 of 15 2.32 1.77 0.00012 DOID:2531 Hematologic cancer 6 of 190 1.52 1.71 1.11e-05 DOID:684 Hepatocellular carcinoma 3 of 17 2.26 1.7 0.00016 DOID:219 Colon cancer 3 of 17 2.26 1.7 0.00016 DOID:1795 Tumor of exocrine pancreas 3 of 17 2.26 1.7 0.00016 DOID:1612 Breast cancer 4 of 62 1.83 1.65 9.71e-05 DOID:3908 Lung non-small cell carcinoma 3 of 26 2.08 1.51 0.00037 DOID:11166 Obsolete papillomavirus infectious disease 2 of 2 3.02 1.5 0.00062 DOID:0050615 Respiratory system cancer 4 of 81 1.71 1.49 0.00019 DOID:170 Endocrine gland cancer 4 of 93 1.65 1.4 0.00028 DOID:0050686 Organ system cancer 10 of 757 1.14 1.37 1.15e-06 DOID:9261 Nasopharynx carcinoma 2 of 4 2.71 1.33 0.0014 DOID:1240 Leukemia 4 of 104 1.6 1.32 0.00041 DOID:8584 Burkitt lymphoma 2 of 5 2.62 1.27 0.0018 DOID:12704 Ataxia telangiectasia 2 of 5 2.62 1.27 0.0018 DOID:162 Cancer 11 of 978 1.07 1.26 1.15e-06 DOID:1909 Melanoma 3 of 46 1.83 1.22 0.0014 DOID:3498 Pancreatic ductal adenocarcinoma 2 of 6 2.54 1.22 0.0023 DOID:3012 Li-Fraumeni syndrome 2 of 6 2.54 1.22 0.0023 DOID:8557 Oropharynx cancer 2 of 7 2.47 1.18 0.0028 DOID:1037 Lymphoid leukemia 3 of 53 1.77 1.15 0.0019 DOID:2893 Cervix carcinoma 2 of 9 2.36 1.12 0.0036 DOID:3347 Osteosarcoma 2 of 10 2.32 1.09 0.0041 DOID:0060108 Brain glioma 2 of 10 2.32 1.09 0.0041 DOID:4159 Skin cancer 3 of 63 1.69 1.07 0.0028 DOID:5520 Head and neck squamous cell carcinoma 2 of 11 2.28 1.06 0.0047 DOID:8923 Skin melanoma 2 of 12 2.24 1.03 0.0054 DOID:0050621 Respiratory system benign neoplasm 2 of 12 2.24 1.03 0.0054 DOID:345 Uterine disease 3 of 72 1.64 1.01 0.0036 DOID:0050687 Cell type cancer 6 of 451 1.14 1.0 0.00041 DOID:305 Carcinoma 5 of 307 1.23 0.99 0.0010 DOID:786 Laryngeal disease 2 of 14 2.17 0.99 0.0068 DOID:3068 Glioblastoma multiforme 2 of 14 2.17 0.99 0.0068 DOID:768 Retinoblastoma 2 of 16 2.11 0.95 0.0080 DOID:4001 Ovarian carcinoma 2 of 16 2.11 0.95 0.0080 DOID:229 Female reproductive system disease 4 of 192 1.33 0.94 0.0028 DOID:0050745 Diffuse large B-cell lymphoma 2 of 17 2.09 0.93 0.0086 DOID:8618 Oral cavity cancer 2 of 18 2.06 0.92 0.0094 DOID:120 Female reproductive organ cancer 3 of 100 1.49 0.86 0.0077 DOID:77 Gastrointestinal system disease 6 of 576 1.03 0.82 0.0014 DOID:9952 Acute lymphoblastic leukemia 2 of 26 1.9 0.8 0.0165 DOID:2513 Basal cell carcinoma 2 of 27 1.89 0.78 0.0176 DOID:3459 Breast carcinoma 2 of 29 1.85 0.76 0.0195 DOID:403 Mouth disease 3 of 130 1.38 0.74 0.0137 DOID:10534 Stomach cancer 2 of 31 1.83 0.74 0.0216 DOID:289 Endometriosis 2 of 34 1.79 0.71 0.0252 DOID:0070004 Myeloid neoplasm 2 of 38 1.74 0.68 0.0294 DOID:6713 Cerebrovascular disease 2 of 46 1.65 0.61 0.0400 DOID:28 Endocrine system disease 4 of 398 1.02 0.55 0.0243 DOID:225 Syndrome 6 of 1214 0.71 0.41 0.0315 DOID:7 Disease of anatomical entity 12 of 4798 0.41 0.31 0.0185 """ # Function to parse the raw data into a structured format def parse_data(data): """ Parse the raw data string into a list of lists containing numerical values and disease names. Args: data (str): Raw data string with lines containing DOID, disease name, successes of total, and values. Returns: list: List of lists, each containing [successes, total, value1, value2, p_value, disease]. """ lines = data.strip().split('\n') data_list = [] for line in lines: parts = line.split() # Split on any whitespace # Find the last occurrence of 'of' by searching from the end i = parts[::-1].index('of') # Find 'of' from the reversed list i = len(parts) - i - 1 # Convert to original index disease = ' '.join(parts[1:i-1]) # Disease name from after DOID to before successes successes = int(parts[i-1]) # Number before 'of' total = int(parts[i+1]) # Number after 'of' value1 = float(parts[i+2]) # First number after total value2 = float(parts[i+3]) # Second number p_value = float(parts[i+4]) # Third number (p-value) data_list.append([successes, total, value1, value2, p_value, disease]) return data_list # Parse data into a DataFrame data_list = parse_data(data) df = pd.DataFrame(data_list, columns=['Successes', 'Total', 'Value1', 'Value2', 'P_Value', 'Disease']) # Feature Engineering df['Accuracy'] = df['Successes'] / df['Total'] # Success rate df['LogTotal'] = np.log1p(df['Total']) # Log of total cases df['LogPValue'] = -np.log10(df['P_Value']) # Negative log10 of p-value # Define cancer-related terms for labeling cancer_terms = ["cancer", "carcinoma", "melanoma", "lymphoma", "leukemia", "sarcoma", "glioma", "adenocarcinoma", "retinoblastoma", "glioblastoma"] # Create target variable (1 for cancer, 0 for non-cancer) y = np.array([1 if any(term in disease.lower() for term in cancer_terms) else 0 for disease in df['Disease']]) # Define feature set X = df[['Successes', 'Total', 'Value1', 'Value2', 'P_Value', 'Accuracy', 'LogTotal', 'LogPValue']] # Split data into training and testing sets X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42, stratify=y) # Custom F1 scorer to handle zero division def safe_f1(y_true, y_pred): return f1_score(y_true, y_pred, average='binary', zero_division=0) custom_f1 = make_scorer(safe_f1) # --- Model Pipelines and Tuning --- # SVM Pipeline def create_svm_pipeline(): return Pipeline([ ('imputer', SimpleImputer(strategy='median')), ('scaler', StandardScaler()), ('svm', SVC(random_state=42, class_weight='balanced')) ]) def tune_svm(X_train, y_train): param_grid = { 'svm__C': [0.1, 1, 10], 'svm__kernel': ['rbf', 'linear'], 'svm__gamma': ['scale', 'auto', 0.1, 1] } grid_search = GridSearchCV(create_svm_pipeline(), param_grid, scoring=custom_f1, cv=KFold(n_splits=10, shuffle=True, random_state=42), n_jobs=-1) grid_search.fit(X_train, y_train) return grid_search.best_estimator_ # Random Forest Pipeline def create_rf_pipeline(): return Pipeline([ ('imputer', SimpleImputer(strategy='median')), ('scaler', StandardScaler()), ('rf', RandomForestClassifier(random_state=42, class_weight='balanced')) ]) def tune_rf(X_train, y_train): param_grid = { 'rf__n_estimators': [50, 100, 200], 'rf__max_depth': [3, 5, 7], 'rf__min_samples_leaf': [1, 5, 10] } grid_search = GridSearchCV(create_rf_pipeline(), param_grid, scoring=custom_f1, cv=KFold(n_splits=10, shuffle=True, random_state=42), n_jobs=-1) grid_search.fit(X_train, y_train) return grid_search.best_estimator_ # Gradient Boosting Pipeline def create_gb_pipeline(): return Pipeline([ ('imputer', SimpleImputer(strategy='median')), ('scaler', StandardScaler()), ('gb', GradientBoostingClassifier(random_state=42)) ]) def tune_gb(X_train, y_train): param_grid = { 'gb__n_estimators': [50, 100, 200], 'gb__learning_rate': [0.01, 0.1, 0.2], 'gb__max_depth': [3, 5, 7] } grid_search = GridSearchCV(create_gb_pipeline(), param_grid, scoring=custom_f1, cv=KFold(n_splits=10, shuffle=True, random_state=42), n_jobs=-1) grid_search.fit(X_train, y_train) return grid_search.best_estimator_ # --- Training and Evaluation Function --- def train_evaluate_model(model_name, create_pipeline_func, tune_func, X_train, y_train, X_test, y_test): """ Train and evaluate a model using the specified pipeline and tuning function. Args: model_name (str): Name of the model (e.g., "SVM"). create_pipeline_func (callable): Function to create the model pipeline. tune_func (callable): Function to tune the model with GridSearchCV. X_train, y_train: Training data and labels. X_test, y_test: Testing data and labels. Returns: best_model: The trained and tuned model. """ print(f"\nTraining and Evaluating {model_name}...") best_model = tune_func(X_train, y_train) y_pred = best_model.predict(X_test) accuracy = accuracy_score(y_test, y_pred) f1 = safe_f1(y_test, y_pred) print(f"{model_name} Accuracy on Test Set: {accuracy:.4f}") print(f"{model_name} F1 Score on Test Set: {f1:.4f}") print(f"{model_name} Classification Report:\n{classification_report(y_test, y_pred)}") return best_model # --- Train and Evaluate Models --- svm_model = train_evaluate_model("SVM", create_svm_pipeline, tune_svm, X_train, y_train, X_test, y_test) rf_model = train_evaluate_model("Random Forest", create_rf_pipeline, tune_rf, X_train, y_train, X_test, y_test) gb_model = train_evaluate_model("Gradient Boosting", create_gb_pipeline, tune_gb, X_train, y_train, X_test, y_test)

OUTPUT:

Training and Evaluating SVM... SVM Accuracy on Test Set: 0.7500 SVM F1 Score on Test Set: 0.8421 SVM Classification Report: precision recall f1-score support 0 0.50 0.33 0.40 3 1 0.80 0.89 0.84 9 accuracy 0.75 12 macro avg 0.65 0.61 0.62 12 weighted avg 0.72 0.75 0.73 12 Training and Evaluating Random Forest... Random Forest Accuracy on Test Set: 0.7500 Random Forest F1 Score on Test Set: 0.8421 Random Forest Classification Report: precision recall f1-score support 0 0.50 0.33 0.40 3 1 0.80 0.89 0.84 9 accuracy 0.75 12 macro avg 0.65 0.61 0.62 12 weighted avg 0.72 0.75 0.73 12 Training and Evaluating Gradient Boosting... Gradient Boosting Accuracy on Test Set: 0.7500 Gradient Boosting F1 Score on Test Set: 0.8421 Gradient Boosting Classification Report: precision recall f1-score support 0 0.50 0.33 0.40 3 1 0.80 0.89 0.84 9 accuracy 0.75 12 macro avg 0.65 0.61 0.62 12 weighted avg 0.72 0.75 0.73 12

EXTENDED CODE:

import pandas as pd import numpy as np from sklearn.model_selection import train_test_split, KFold, GridSearchCV from sklearn.preprocessing import StandardScaler from sklearn.svm import SVC from sklearn.ensemble import RandomForestClassifier, GradientBoostingClassifier from sklearn.metrics import accuracy_score, classification_report, make_scorer, f1_score from sklearn.pipeline import Pipeline from sklearn.impute import SimpleImputer import warnings import plotly.express as px # For dynamic plotting warnings.filterwarnings("ignore") # Raw data string (disease dataset) data = """ DOID:11054 Urinary bladder cancer 5 of 38 2.13 2.53 1.15e-06 DOID:5672 Large intestine cancer 5 of 44 2.07 2.49 1.15e-06 DOID:5093 Thoracic cancer 5 of 71 1.86 2.13 4.77e-06 DOID:9256 Colorectal cancer 4 of 32 2.11 2.1 1.36e-05 DOID:707 B-cell lymphoma 4 of 34 2.09 2.07 1.57e-05 DOID:9561 Nasopharyngeal disease 3 of 9 2.54 1.99 4.49e-05 DOID:0060119 Pharynx cancer 3 of 12 2.41 1.84 8.75e-05 DOID:0060058 Lymphoma 5 of 105 1.69 1.82 1.77e-05 DOID:3910 Lung adenocarcinoma 3 of 13 2.38 1.82 9.71e-05 DOID:4905 Pancreatic carcinoma 3 of 15 2.32 1.77 0.00012 DOID:2531 Hematologic cancer 6 of 190 1.52 1.71 1.11e-05 DOID:684 Hepatocellular carcinoma 3 of 17 2.26 1.7 0.00016 DOID:219 Colon cancer 3 of 17 2.26 1.7 0.00016 DOID:1795 Tumor of exocrine pancreas 3 of 17 2.26 1.7 0.00016 DOID:1612 Breast cancer 4 of 62 1.83 1.65 9.71e-05 DOID:3908 Lung non-small cell carcinoma 3 of 26 2.08 1.51 0.00037 DOID:11166 Obsolete papillomavirus infectious disease 2 of 2 3.02 1.5 0.00062 DOID:0050615 Respiratory system cancer 4 of 81 1.71 1.49 0.00019 DOID:170 Endocrine gland cancer 4 of 93 1.65 1.4 0.00028 DOID:0050686 Organ system cancer 10 of 757 1.14 1.37 1.15e-06 DOID:9261 Nasopharynx carcinoma 2 of 4 2.71 1.33 0.0014 DOID:1240 Leukemia 4 of 104 1.6 1.32 0.00041 DOID:8584 Burkitt lymphoma 2 of 5 2.62 1.27 0.0018 DOID:12704 Ataxia telangiectasia 2 of 5 2.62 1.27 0.0018 DOID:162 Cancer 11 of 978 1.07 1.26 1.15e-06 DOID:1909 Melanoma 3 of 46 1.83 1.22 0.0014 DOID:3498 Pancreatic ductal adenocarcinoma 2 of 6 2.54 1.22 0.0023 DOID:3012 Li-Fraumeni syndrome 2 of 6 2.54 1.22 0.0023 DOID:8557 Oropharynx cancer 2 of 7 2.47 1.18 0.0028 DOID:1037 Lymphoid leukemia 3 of 53 1.77 1.15 0.0019 DOID:2893 Cervix carcinoma 2 of 9 2.36 1.12 0.0036 DOID:3347 Osteosarcoma 2 of 10 2.32 1.09 0.0041 DOID:0060108 Brain glioma 2 of 10 2.32 1.09 0.0041 DOID:4159 Skin cancer 3 of 63 1.69 1.07 0.0028 DOID:5520 Head and neck squamous cell carcinoma 2 of 11 2.28 1.06 0.0047 DOID:8923 Skin melanoma 2 of 12 2.24 1.03 0.0054 DOID:0050621 Respiratory system benign neoplasm 2 of 12 2.24 1.03 0.0054 DOID:345 Uterine disease 3 of 72 1.64 1.01 0.0036 DOID:0050687 Cell type cancer 6 of 451 1.14 1.0 0.00041 DOID:305 Carcinoma 5 of 307 1.23 0.99 0.0010 DOID:786 Laryngeal disease 2 of 14 2.17 0.99 0.0068 DOID:3068 Glioblastoma multiforme 2 of 14 2.17 0.99 0.0068 DOID:768 Retinoblastoma 2 of 16 2.11 0.95 0.0080 DOID:4001 Ovarian carcinoma 2 of 16 2.11 0.95 0.0080 DOID:229 Female reproductive system disease 4 of 192 1.33 0.94 0.0028 DOID:0050745 Diffuse large B-cell lymphoma 2 of 17 2.09 0.93 0.0086 DOID:8618 Oral cavity cancer 2 of 18 2.06 0.92 0.0094 DOID:120 Female reproductive organ cancer 3 of 100 1.49 0.86 0.0077 DOID:77 Gastrointestinal system disease 6 of 576 1.03 0.82 0.0014 DOID:9952 Acute lymphoblastic leukemia 2 of 26 1.9 0.8 0.0165 DOID:2513 Basal cell carcinoma 2 of 27 1.89 0.78 0.0176 DOID:3459 Breast carcinoma 2 of 29 1.85 0.76 0.0195 DOID:403 Mouth disease 3 of 130 1.38 0.74 0.0137 DOID:10534 Stomach cancer 2 of 31 1.83 0.74 0.0216 DOID:289 Endometriosis 2 of 34 1.79 0.71 0.0252 DOID:0070004 Myeloid neoplasm 2 of 38 1.74 0.68 0.0294 DOID:6713 Cerebrovascular disease 2 of 46 1.65 0.61 0.0400 DOID:28 Endocrine system disease 4 of 398 1.02 0.55 0.0243 DOID:225 Syndrome 6 of 1214 0.71 0.41 0.0315 DOID:7 Disease of anatomical entity 12 of 4798 0.41 0.31 0.0185 """ # Parse data into a structured format def parse_data(data): lines = data.strip().split('\n') data_list = [] for line in lines: parts = line.split() i = len(parts) - parts[::-1].index('of') - 1 # Find last 'of' disease = ' '.join(parts[1:i-1]) successes = int(parts[i-1]) total = int(parts[i+1]) value1 = float(parts[i+2]) value2 = float(parts[i+3]) p_value = float(parts[i+4]) data_list.append([successes, total, value1, value2, p_value, disease]) return data_list # Create DataFrame data_list = parse_data(data) df = pd.DataFrame(data_list, columns=['Successes', 'Total', 'Value1', 'Value2', 'P_Value', 'Disease']) # Feature Engineering df['Accuracy'] = df['Successes'] / df['Total'] df['LogTotal'] = np.log1p(df['Total']) df['LogPValue'] = -np.log10(df['P_Value']) # Define target variable (cancer vs. non-cancer) cancer_terms = ["cancer", "carcinoma", "melanoma", "lymphoma", "leukemia", "sarcoma", "glioma", "adenocarcinoma", "retinoblastoma", "glioblastoma"] y = np.array([1 if any(term in disease.lower() for term in cancer_terms) else 0 for disease in df['Disease']]) # Features X = df[['Successes', 'Total', 'Value1', 'Value2', 'P_Value', 'Accuracy', 'LogTotal', 'LogPValue']] # Split data X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42, stratify=y) # Custom F1 scorer def safe_f1(y_true, y_pred): return f1_score(y_true, y_pred, average='binary', zero_division=0) custom_f1 = make_scorer(safe_f1) # Model Pipelines def create_svm_pipeline(): return Pipeline([ ('imputer', SimpleImputer(strategy='median')), ('scaler', StandardScaler()), ('svm', SVC(random_state=42, class_weight='balanced')) ]) def tune_svm(X_train, y_train): param_grid = {'svm__C': [0.1, 1, 10], 'svm__kernel': ['rbf', 'linear'], 'svm__gamma': ['scale', 'auto', 0.1, 1]} grid_search = GridSearchCV(create_svm_pipeline(), param_grid, scoring=custom_f1, cv=KFold(n_splits=10, shuffle=True, random_state=42), n_jobs=-1) grid_search.fit(X_train, y_train) return grid_search.best_estimator_ def create_rf_pipeline(): return Pipeline([ ('imputer', SimpleImputer(strategy='median')), ('scaler', StandardScaler()), ('rf', RandomForestClassifier(random_state=42, class_weight='balanced')) ]) def tune_rf(X_train, y_train): param_grid = {'rf__n_estimators': [50, 100, 200], 'rf__max_depth': [3, 5, 7], 'rf__min_samples_leaf': [1, 5, 10]} grid_search = GridSearchCV(create_rf_pipeline(), param_grid, scoring=custom_f1, cv=KFold(n_splits=10, shuffle=True, random_state=42), n_jobs=-1) grid_search.fit(X_train, y_train) return grid_search.best_estimator_ def create_gb_pipeline(): return Pipeline([ ('imputer', SimpleImputer(strategy='median')), ('scaler', StandardScaler()), ('gb', GradientBoostingClassifier(random_state=42)) ]) def tune_gb(X_train, y_train): param_grid = {'gb__n_estimators': [50, 100, 200], 'gb__learning_rate': [0.01, 0.1, 0.2], 'gb__max_depth': [3, 5, 7]} grid_search = GridSearchCV(create_gb_pipeline(), param_grid, scoring=custom_f1, cv=KFold(n_splits=10, shuffle=True, random_state=42), n_jobs=-1) grid_search.fit(X_train, y_train) return grid_search.best_estimator_ # Train and Evaluate Models def train_evaluate_model(model_name, create_func, tune_func, X_train, y_train, X_test, y_test): print(f"\nTraining and Evaluating {model_name}...") best_model = tune_func(X_train, y_train) y_pred = best_model.predict(X_test) accuracy = accuracy_score(y_test, y_pred) f1 = safe_f1(y_test, y_pred) print(f"{model_name} Accuracy: {accuracy:.4f}") print(f"{model_name} F1 Score: {f1:.4f}") print(f"{model_name} Classification Report:\n{classification_report(y_test, y_pred)}") return best_model, {'Model': model_name, 'Accuracy': accuracy, 'F1 Score': f1} # Train models and collect metrics models = ['SVM', 'Random Forest', 'Gradient Boosting'] create_funcs = [create_svm_pipeline, create_rf_pipeline, create_gb_pipeline] tune_funcs = [tune_svm, tune_rf, tune_gb] metrics_list = [] for model_name, create_func, tune_func in zip(models, create_funcs, tune_funcs): _, metrics = train_evaluate_model(model_name, create_func, tune_func, X_train, y_train, X_test, y_test) metrics_list.append(metrics) # Create DataFrame for plotting metrics_df = pd.DataFrame(metrics_list) # Generate Dynamic Plot fig = px.bar(metrics_df, x='Model', y=['Accuracy', 'F1 Score'], barmode='group', title='Model Performance Comparison', labels={'value': 'Score', 'variable': 'Metric'}, height=500) fig.update_layout(showlegend=True) fig.show()

OUTPUT:

Training and Evaluating Random Forest... Random Forest Accuracy: 0.7500 Random Forest F1 Score: 0.8421 Random Forest Classification Report: precision recall f1-score support 0 0.50 0.33 0.40 3 1 0.80 0.89 0.84 9 accuracy 0.75 12 macro avg 0.65 0.61 0.62 12 weighted avg 0.72 0.75 0.73 12 Training and Evaluating Gradient Boosting... Gradient Boosting Accuracy: 0.7500 Gradient Boosting F1 Score: 0.8421 Gradient Boosting Classification Report: precision recall f1-score support 0 0.50 0.33 0.40 3 1 0.80 0.89 0.84 9 accuracy 0.75 12 macro avg 0.65 0.61 0.62 12 weighted avg 0.72 0.75 0.73 12

Disease-gene association code © 2025 by Devise Foundation is licensed under CC BY-NC-ND 4.0