3.3. Field-Based QSAR Model

The 3D-QSAR analysis using field-based methods was performed with the QSAR tool of the Schrodinger Suite. The 3D-QSAR method builds the model by relating the known activities and molecular elements of the training set using the OPLS_2005 force field. The steric and electrostatic field around the ligand on a 3D grid was calculated using the field-based 3D-QSAR. The force-field-based QSAR model is an alignment-dependent method in which the interaction energy terms of the molecular field are correlated with biological activities using multivariate statistical analysis. In the 3D-QSAR model based on Gaussian force fields, interaction energy calculations were performed using steric, electrostatic, hydrogen bond donor (HBD), and hydrogen bond acceptor (HBA) potential fields using Gaussian equations for the field calculations. The fields were calculated on an orthohedral grid enclosing the training set molecules, with a spacing of 1 Å and extending 3 Å beyond the boundaries of this set. The threshold for van der Waals and electrostatic interactions was set at 30 kcal/mol, eliminating points closer than 2 Å from any of the atoms in the training set. During the PLS procedure, all variables (grid points) with a standard deviation less than 0.05 were removed.

The lattice and probe step sizes were adjusted automatically. The partial least squares (PLS) analysis is applied to construct the best model through the linear correlation of FFQSAR and GFQSAR concerning pEC₅₀ [18,22,23]. The maximum number of PLS factors was set to 5. A cross-validation analysis was performed using the leave-one-out method.

The external predictive ability of each model constructed was assessed by calculating the predictive correlation coefficient (R²_test). In addition, the models were also subjected to external validation criteria according to the test proposed by Golbraikh and Tropsha (see Supplementary Material in Section S2.4) [39,40]. All these calculations were carried out with the DTC Lab software tools (https://dtclab.webs.com/software-tools, accessed on 25 March 2021).