Prediction of Antimalarial Drug-Decorated Nanoparticle Delivery Systems with Random Forest Models

Area of Science:

• Medicinal Chemistry
• Computational Biology
• Nanotechnology

Background:

• Drug-decorated nanoparticles (DDNPs) show promise in medical applications, particularly for treating diseases like malaria.
• Predicting the efficacy of DDNPs requires extensive experimental screening, which is time-consuming and resource-intensive.

Purpose of the Study:

• To develop and validate a computational model for predicting the antimalarial activity of DDNPs.
• To reduce the need for experimental screening by employing virtual screening methods.
• To leverage information fusion of experimental data and molecular features for enhanced predictive accuracy.

Main Methods:

• Combined Perturbation Theory with Machine Learning and Information Fusion (PTMLIF) for model development.
• Utilized fused experimental data from nanoparticle assays and the ChEMBL database for compound information.
• Engineered experiment-centered features representing drug/compound and nanoparticle perturbations.
• Trained and evaluated eight machine learning classifiers on a dataset of 249,992 examples with 107 input features.

Main Results:

• The Random Forest classifier achieved the highest performance, utilizing 27 selected features.
• The model demonstrated high predictive accuracy with a mean Area Under the ROC Curve (AUC) of 0.9921 ± 0.000244 via 10-fold cross-validation.
• Information fusion of experiment-centered features proved powerful for predicting antimalarial activity.

Conclusions:

• The PTMLIF approach offers a robust and efficient method for virtual screening of DDNPs against malaria.
• This computational strategy can significantly accelerate the discovery and development of new antimalarial therapies.
• The study highlights the potential of integrating diverse data sources and advanced machine learning for drug discovery.