A QSAR Study Based on SVM for the Compound of Hydroxyl Benzoic Esters

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Cheminformatics

Background:

• Hydroxyl benzoic esters are widely utilized as preservatives in food, pharmaceuticals, and cosmetics.
• Understanding the relationship between molecular structure and antibacterial activity is crucial for developing effective preservatives.

Purpose of the Study:

• To build Quantitative Structure-Activity Relationship (QSAR) models for 25 hydroxyl benzoic esters.
• To explore the correlation between molecular structure and antibacterial activity.
• To predict the antimicrobial properties of structurally similar compounds.

Main Methods:

• Utilized R language to construct QSAR models based on support vector machine (SVM).
• Incorporated quantum chemical parameters and molecular connectivity indices as descriptors.
• Evaluated model performance using External Standard Deviation Error of Prediction (SDEPext), R², and leave-one-out cross-validation (Q²LOO).

Main Results:

• Four nonlinear models demonstrated R² and Q²LOO values exceeding 0.6, with SDEPext ranging from 0.189 to 0.222.
• The developed SVM models outperformed a multiple linear regression (MLR) model (R² = 0.421).
• Models exhibited good reliability, stability, robustness, and external predictive ability, especially the linear kernel function and eps-regression type.

Conclusions:

• QSAR models effectively predict the antimicrobial activity of hydroxyl benzoic esters.
• The developed models provide a valuable tool for designing novel compounds with enhanced preservative properties.
• The study highlights the potential of computational approaches in accelerating the discovery of new antimicrobial agents.