QSAR models for isoindolinone-based p53-MDM2 interaction inhibitors using linear and non-linear statistical methods.

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• The p53-MDM2 protein-protein interaction is a key target for developing novel anticancer therapies.
• Isoindolinone derivatives have shown promise as inhibitors of this interaction.
• Developing predictive models is crucial for optimizing these inhibitors.

Purpose of the Study:

• To systematically investigate the quantitative structure-activity relationships (QSAR) of isoindolinone-based p53-MDM2 interaction inhibitors.
• To develop and compare the performance of linear and non-linear QSAR models.
• To validate the best performing models for virtual screening applications.

Main Methods:

• Systematic 2D-QSAR studies were performed on a dataset of 98 isoindolinone-based compounds.
• Linear methods, including forward stepwise-multiple linear regression (FS-MLR), were employed.
• Non-linear methods, such as enhanced replacement method-multiple linear regression (ERM-MLR) and support vector machine regression (SVMR), were utilized.
• Model performance was evaluated using training, leave-one-out (LOO), and test set parameters.
• Receiver operating characteristic (ROC) studies were conducted for virtual screening validation.

Main Results:

• The FS-MLR model achieved good statistical performance (R(2)(train)=0.881, Q(2)(loo)=0.847, R(2)(test)=0.854).
• More accurate models were developed using ERM-MLR (R(2)(train)=0.914, Q(2)(loo)=0.894, R(2)(test)=0.903) and SVMR (R(2)(train)=0.924, Q(2)(loo)=0.920, R(test)(2)=0.874).
• The ERM-MLR and SVMR models demonstrated reliability and applicability in virtual screening, confirmed by ROC analyses.

Conclusions:

• Advanced QSAR modeling, particularly ERM-MLR and SVMR, can accurately predict the inhibitory activity of isoindolinone-based compounds against the p53-MDM2 interaction.
• These validated models are valuable tools for the rational design and optimization of new anticancer agents.
• The study highlights the potential of computational approaches in accelerating drug discovery for cancer therapy.