Machine learning-based analysis on pharmaceutical compounds interaction with polymer to estimate drug solubility in...

Area of Science:

• Computational chemistry and cheminformatics
• Machine learning applications in drug discovery
• Pharmaceutical formulation development

Background:

• Accurate prediction of drug solubility and activity is crucial for effective pharmaceutical formulation.
• Traditional methods for determining these properties can be time-consuming and resource-intensive.
• Machine learning offers a promising approach to accelerate and enhance prediction accuracy.

Purpose of the Study:

• To introduce a sophisticated machine learning framework for predicting drug solubility and activity in formulations.
• To leverage ensemble learning techniques to improve prediction accuracy.
• To optimize the predictive model through advanced feature selection and hyperparameter tuning.

Main Methods:

• Utilized a comprehensive dataset of over 12,000 data rows and 24 input features.
• Employed ensemble learning by improving base models (Decision Tree, K-Nearest Neighbors, Multilayer Perceptron) with AdaBoost.
• Implemented Recursive Feature Elimination for feature selection and Harmony Search for hyperparameter optimization.

Main Results:

• The ADA-DT model achieved an R² score of 0.9738 for drug solubility prediction.
• The ADA-KNN model achieved an R² value of 0.9545 for gamma prediction.
• Both models demonstrated high accuracy with low Mean Squared Error (MSE) and Mean Absolute Error (MAE).

Conclusions:

• Ensemble learning, combined with advanced feature selection and hyperparameter optimization, accurately predicts complex biochemical properties.
• The developed framework provides a powerful tool for enhancing drug formulation development.
• This approach can significantly accelerate the drug discovery and development pipeline.