Interpretation of CoMFA Results – A Probe Set Study Using Hydrophobic Fields

The main purpose of the study was to examine the correctness of interpretation of CoMFA results using an artificially designed data set with predefined contributions of the compound substituents to activity. The activity values were assigned according to additive (sum of substituent π-constants) and nonlinear (sum of absolute values of substituent π-constants) dependencies on hydrophobicity. Predictions by 3D (HINT hydrophobic fields) and logP (HINT and CLOGP values) presentations of hydrophobicity were compared and similarity between standard CoMFA and hydrophobic fields was evaluated. The main results are: (i) the cross- validated R2 (Q2) values with the first PLS components of the field models may be indicative for identification of the underlying property(ies) in the data set providing each field alone yields a satisfactory predictive model with several components; (ii) the logP values alone are not predictive when the target property is nonlinearly dependent on the explanatory property; (iii) similarity between fields of different nature (steric, electrostatic, hydrophobic) can be evaluated by cross- and non-cross-validation correlations between the X-scores of the first components – low Q2 and R2 suggest that the most informative variances of the compared fields are different; (iv) the CoMFA graphical display is very much dependent on the distribution of the positive and negative field terms – if their contributions to the whole field signal are not symmetrically distributed the default CoMFA contour view setting can lead to wrong displays and, consequently, wrong interpretation of the results.