Prediction and Screening Model for Products Based on Fusion Regression and XGBoost Classification

Area of Science:

• Computational chemistry and cheminformatics
• Drug discovery and development
• Machine learning in pharmacology

Background:

• Drug discovery failures often stem from low bioactivity and poor absorption, distribution, metabolism, excretion, and toxicity (ADMET) profiles of candidate compounds.
• Accurate prediction of bioactivity and ADMET properties early in the drug discovery process is crucial for success.
• Systematic evaluation of these properties is necessary to mitigate risks and reduce attrition rates.

Purpose of the Study:

• To propose a data-driven model for screening pharmaceutical product candidates with enhanced bioactivity and ADMET properties.
• To develop a quantitative prediction method for bioactivity value and a robust prediction method for ADMET properties.
• To define a screening method (BVAP) based on predicted values for efficient candidate selection.

Main Methods:

• Quantitative prediction of bioactivity value using fused regression of Light Gradient Boosting Machine (LGBM) and backpropagation neural network (BP-NN).
• Prediction of ADMET properties utilizing the XGBoost algorithm.
• Development of the Bioactivity Value and ADMET Properties (BVAP) method for screening drug candidates based on predicted performance.

Main Results:

• The fusion regression model achieved a Mean Square Error (MSE) of 1.1496 for bioactivity prediction.
• XGBoost demonstrated high prediction accuracy for ADMET properties: Caco-2 (94.0%), CYP3A4 (95.7%), HERG (89.4%), hob (88.6%), and Mn (96.2%).
• The proposed XGBoost method outperformed traditional methods (SVM, RF, KNN, LDA, NB) in prediction accuracy and AUC for ADMET properties.

Conclusions:

• The data-driven screening model effectively predicts bioactivity and ADMET properties, aiding in the selection of promising drug candidates.
• The XGBoost algorithm shows superior performance in predicting ADMET properties, offering significant guiding value for pharmaceutical development.
• This approach enhances the efficiency and success rate of pharmaceutical product development by identifying candidates with desirable bioactivity, pharmacokinetic, and safety profiles.