A new model of flavonoids affinity towards P-glycoprotein: genetic algorithm-support vector machine with features...

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Pharmacology

Background:

• Flavonoids demonstrate significant binding affinity for the cytosolic nucleotide-binding domain of P-glycoprotein (P-gp).
• Understanding these interactions is crucial for developing targeted therapies and predicting drug efficacy.
• P-gp plays a vital role in multidrug resistance, making its inhibition a key area of research.

Purpose of the Study:

• To develop robust quantitative structure-activity relationship (QSAR) models for predicting flavonoid affinity to P-gp.
• To explore novel computational methods for optimizing QSAR model performance.
• To establish a reliable predictive model for flavonoid-P-gp interactions.

Main Methods:

• Support vector machines (SVMs) were employed for QSAR modeling.
• A hybrid optimization approach combining particle swarm optimization with genetic algorithms was utilized to optimize SVM kernel parameters and feature selection.
• DRAGON descriptors were used to represent flavonoid compounds.

Main Results:

• The developed QSAR model achieved a high correlation coefficient (R ²) of 0.924 for the training set and 0.941 for the external validation set.
• Low root-mean-square error (RMSE) values (0.0588 for training, 0.0443 for cross-validation) indicate high accuracy.
• A mean Q ² value of 0.824 from leave-many-out cross-validation demonstrates the model's stability and predictive power.

Conclusions:

• The novel computational approach provides a stable and accurate model for predicting flavonoid affinity to P-gp.
• The model exhibits excellent predictive ability and a well-defined applicability domain.
• This research offers valuable insights for the rational design of flavonoids targeting P-gp.