Research in Medicine

Background and aim: Due to the complexity of drug discovery processes and the high cost of biological testing, computational methods in medicinal chemistry have gained attention as effective tools for streamlining drug design. One such approach is the study of Quantitative Structure–Activity Relationships (QSAR), which utilizes mathematical and statistical models to predict the biological activity of chemical compounds based on their structural features. By extracting molecular descriptors and establishing a mathematical equation between chemical structure and biological activity, QSAR enables the prediction of biological behavior without the need for synthesis and experimental evaluation. This strategy not only reduces time and cost but also plays a significant role in optimizing drug structures and enhancing their efficacy.
In the present study, QSAR analysis was performed on a series of monoketo acid derivatives previously reported to exhibit inhibitory activity against HIV. The resulting model, with strong predictive capability, can be used for the rational design of new compounds with potential anti-HIV activity and serves as a valuable tool in the development of HIV inhibitors.
Materials and methods: This study was conducted using computational techniques. Initially, the chemical structures of the selected compounds were drawn and their energy levels optimized. Molecular descriptors were then calculated using Dragon and PaDEL software. Constant, near-constant, and highly correlated descriptors were removed to improve model quality. The dataset was randomly divided into training and test sets. Important descriptors were selected using the stepwise method, and a multiple linear regression (MLR) model was constructed. For model validation, both internal and external validation techniques were applied. The coefficient of determination (R²) for the training and test sets, as well as the Q² LOO (Leave-One-Out cross-validation), were calculated. Acceptable models were defined by R² and Q² values greater than 0.6.
Results: The coefficient of determination (R²) for the training and test sets was 0.834 and 0.808, respectively. The cross-validation coefficient (Q²_LOO) was 0.681, indicating good internal predictivity. The selected descriptors in the final model included R7e, RDF155e, AATS6s, and maxaaCH, belonging to the GETAWAY, RDF, AC, and Electrotopological classes, respectively.
Conclusion: The developed QSAR model appears to be statistically significant and suitable for predicting the inhibitory activity of novel monoketo acid derivatives against HIV-1. This model can be considered a valuable computational tool for the rational design of new anti-HIV agents.