other:Article Title: Structural Analysis and Prediction of Hematotoxicity Using Deep Learning Approaches.
Article Snippet: Important Parameters Used in QSAR Modeling Hyperparameter Software RF MOE2d the number of trees = 2200 KNIME RDKit-d the number of trees = 2400 KNIME MACCS the number of trees = 2800 KNIME CATS the number of trees = 2800 KNIME Estate the number of trees = 600 KNIME ECFP4 the number of trees = 1200 KNIME FCFP4 the number of trees = 2000 KNIME RDKit-f the number of trees = 2000 KNIME XGBoost MOE2d step size shrinkage = 0.1, maximum depth of a tree = 6, the max number of iterations = 1800 KNIME RDKit-d step size shrinkage = 0.2, maximum depth of a tree = 6, the max number of iterations = 1800 KNIME MACCS step size shrinkage = 0.2, maximum depth of a tree = 4, the max number of iterations = 2200 KNIME CATS step size shrinkage = 0.3, maximum depth of a tree = 5, the max number of iterations = 1400 KNIME Estate step size shrinkage = 0.4, maximum depth of a tree = 6, the max number of iterations = 1000 KNIME ECFP4 step size shrinkage = 0.6, maximum depth of a tree = 5, the max number of iterations = 1400 KNIME FCFP4 step size shrinkage = 0.4, maximum depth of a tree = 6, the max number of iterations = 1000 KNIME RDKit-f step size shrinkage = 0.1, maximum depth of a tree = 4, the max number of iterations = 1800 KNIME GBDT MOE2d number of models = 1800, learning rate = 0.5 KNIME RDKit-d number of models = 1000, learning rate = 0.6 KNIME MACCS number of models = 2000, learning rate = 0.4 KNIME CATS number of models = 2200, learning rate = 0.5 KNIME S-3 Estate number of models = 3000, learning rate = 0.6 KNIME ECFP4 number of models = 2000, learning rate = 0.2 KNIME FCFP4 number of models = 600, learning rate = 0.2 KNIME RDKit-f number of models = 2600, learning rate = 0.5 KNIME SVM MOE2d sigma = 0.8 KNIME RDKit-d sigma = 0.2 KNIME MACCS sigma = 1.9 KNIME CATS sigma = 0.4 KNIME Estate sigma = 0.4 KNIME ECFP4 sigma = 1.9 KNIME FCFP4 sigma = 1.8 KNIME RDKit-f sigma = 1.5 KNIME Attentive FP Graph n tasks = 2, dropout = 0.2, batch size = 32, epoch = 39 Python MPNN Graph n tasks = 2, num step message passing = 6, batch size = 32, epoch = 39 Python GCN Graph n tasks = 2, hidden feats = [128,128], predictor hidden feats = 10, batch size = 32, dropout = 0.5, epoch = 31 Python S-4 Model performance and applicability domain evaluation BA = 1 2 (SE + SP) Precision = TN TN+FP Recall = TP TP+FN F1 = 2·Precision·Recall Precision+Recall MCC = TP·TN− FP·FN √(TP+FN)(TP+FP)(TN+FN)(TN+FP) ST = S̅ − Zσ where TP is the number of correctly categorized hematotoxic compounds, TN is the number of correctly classified non-hematotoxic compounds, FP represents the number of incorrectly classified hematotoxic compounds, and FN represents the number of incorrectly classified non-hematotoxic compounds.
|
