QSAR modeling for thiolactomycin analogues using genetic algorithm optimized artificial neural networks

Abstract

Fatty acid biosynthesis (FAB) is essential for bacterial survival. Components of this biosynthetic pathway have been identified as attractive targets for the development of new antibacterial agents. Thiolactomycin is an inhibitor of type II fatty acid synthase (FAS). Two-dimension quantitative structure–activity relationship (2D QSAR) such as partial least squares (PLS), quadratic partial least squares (QPLS), artificial neural networks (ANN), genetic algorithm (GA) optimized ANN connection weights (GA-ANN), GA select the most relevant descriptors (GA-ANN-GA) are conducted on a series of potent thiolactomycin analogues. Compare the four methods, the pedictive ability the models is evaluated by the root-mean-square error (RMSE) and R ² for the training set and the test set. The GA-ANN-GA show the best results, the RMSE for the training set and the test set are 0.0718 and 0.9473, respectively, the R ² for the training set and the test set are 0.0702 and 0.9504, respectively. GA optimized ANN will provide a superior alternative method 2D QSAR models.

Keywords:

• FabH inhibitors
• Thiolactomycin
• QSAR
• GA optimized ANN

Acknowledgements

This work was partially supported by China West Medical College.