Use computational approaches to predict the anti- malarial pharmacodynamic property of compounds

• 28 sources (Harvard, GSK, GNF, Novartis, ...)
• Major: ChEMBL-NTD, PubChem
• data: 196,199 instances

• Open-source cheminformatics software RDKit
• Auxiliary features: 196
• Graph-based signature: 425

Used different graph-based signature combinations to generate our features and select the one that can best describing molecule pattern for anti-malarial drugs

• Gradient Boosting Regressor
• Random Forest
• KNN
• Extra Trees Regressor

Random Forest with cut-off 10 and cut-off step 1 graph-based signature combination yielded the best performance for both the performance for 10-fold cross validation and blind test.

graph-based signature: The number of atoms categorized by pharmacophore type within a certain distance with different increasing steps in the molecular graph