In-silico prediction of sweetness using structure-activity relationship models

Area of Science:

• Computational Chemistry
• Medicinal Chemistry
• Drug Discovery

Background:

• Quantitative structure-activity relationship (QSAR) models are valuable for screening molecular candidates.
• Predicting sweetness potency is crucial for designing novel sweeteners.

Purpose of the Study:

• To develop and validate QSAR models for predicting sweetness potency relative to sucrose.
• To offer a computational tool for rapid screening of potential sweetener compounds.

Main Methods:

• Developed QSAR models using a dataset of 487 compounds.
• Employed Genetic Function Approximation and Artificial Neural Network algorithms.
• Split the dataset into 70:30 training and testing sets.

Main Results:

• Achieved high predictive performance with R²test = 0.832 for Genetic Function Approximation.
• Obtained R²test = 0.831 for Artificial Neural Network models.
• Demonstrated the models' capability for accurate sweetness potency prediction.

Conclusions:

• The developed QSAR models provide an efficient method for prescreening molecules.
• These models can guide the design of novel sweeteners by complementing molecular modeling and experimental approaches.