Dihydrofolate reductase inhibitors: a quantitative structure–activity relationship study using 2D-QSAR and 3D-QSAR methods

Abstract

In this work, we study the structure–activity relationship of a series of Dihydrofolate reductase inhibitors by two-dimensional quantitative activity–structure relationship and three-dimensional quantitative activity–structure relationship techniques. The two-dimensional quantitative activity–structure relationship models were developed by using two different types of topological molecular descriptors, PaDEL and Dragon descriptors. The models showed an excellent predictive power, R2 train = 0.916 and R2 val = 0.806 for the PaDEL, and R2 train = 0.952 and R2 val = 0.963 for those obtained with Dragon descriptors. Simple molecular descriptors as maxHCsats, IC3, SPI, SIC2, and GATS5p were adequate to obtain predictive models. The three-dimensional quantitative activity–structure relationship was performed through three variable selected approaches, Partial Linear Square (PLS), Fractional Factorial Design (FFD) and Uninformative Variable Elimination-Partial Linear Square (UVE-PLS) using the Open3DQSAR software. All the 2D and 3D models were validated using two compounds (number 24 and 25), which were synthesized and presented here for the first time. Their biological activities were correctly predicted by all the quantitative activity–structure relationship models. Finally, we proposed three compounds (26, 27, and 28), which showed a high predicted Dihydrofolate reductase inhibitory activity. Molecular docking study suggested that compounds bind to receptor similarly to the most active inhibitors.