| Abstract
| - Displaying an unprecedented structural diversity, 119 I2 ligands, and their pKi values, werecollected and submitted to a comparative molecular fields analysis (CoMFA) study. They werediscerned into three structural subsets (A, B, C), to explore the I2 3D-QSARs from finitestructural systems (A, B, C) to more complex ones (AB, AC, BC, ABC). In addition, variouskey steps of the CoMFA methology were explored. The applied method used two pharmacophoretemplates and seven molecular field combinations (electrostatic, lipophilic, steric), as well aseight alignment methods (two point-by-point and six similarity-based variations). That way,644 CoMFA models were obtained and further selected according to their predictive abilitythrough two filters. The first filter was mainly based on the q2, which internally evaluates thepredictive ability from the training set. For the second filter, the predictive ability was externallyevaluated through the prediction of test sets. Finally, one model was extracted from the wholedata as the best. Indeed, it combines three features of upmost importance for the further designof ligands endowed with high I2 affinity: structural diversity (n = 73), robustness (N = 9, r2 =0.96, s = 0.28, F = 148), and a great fully assessed predictive ability (q2 = 0.50, r2test set = 0.81,ntest set = 46). On the basis of structural data and CoMFA isocontours, some elements of the I2tridimensional pharmacophore are also suggested.
|
