CoMFA comparison with QSAR

Partial Least Squares regression (PLS) is usually performed on a - data matrix to search for a correlation between the thousands of CoMFA descriptors and biological response. However, usually after - variable selection, the PLS model is transformed into and presented as a multiple regression equation to allow comparison with classical QSAR models. 

Robert D, Amat L, Carbo-Dorca R. Quantum similarity QSAR study of inhibitors binding to thrombin, trypsin and factor Xa, including a comparison with CoMFA and CoMSIA methods. Inti J Quantum Chem 2000 80 265-282. 

Comparison of the best 4D-QSAR models with the best CoMFA models revealed that they had similar r2 values, but that the q2 values are higher for the 4D-QSAR models. 

A variety of QSAR models have also been investigated. Comparative molecular field analysis (CoMFA) and comparative molecular similarity indices analysis (CoMSIA) [91, 92] models have been used for the substrate-based and keto acid inhibitors (represented by compounds 27 and 30). Molecular overlays were initially developed for each class, and by comparison of activity data with these overlays, a model can be developed to take into account both steric and electrostatic fields around the inhibitors this is illustrated in Fig. 5 [91]. 

Lopez-Rodriguez ML, Rosado ML, Benhamu B, Morcillo MJ, Fernandez E, Schaper KJ. Synthesis and structure-activity relationships of a new model of arylpiperazines. 2. 3D-QSAR of hydantoin-phenylpiperazine derivatives with affinity for 5-HT1A and alphal receptors. A comparison of CoMFA models. J Med Chem 1997 40 1648-1656. 

A number of questions remain open, including extension of the present approach to objects in 3D space. Are the binary shape codes sufficiently general, i.e., can they be extended to 3D structures We cannot answer these questions since they have not yet been considered. We will, however, outline briefly an approach to the 3D case. A generalization from planar contours to surfaces in a 3D space is possible even if not unique. For example, we first derive for a 3D surface a set of 2D contours. Such are, for example, the electrostatic maps that have found use in QSAR in the well-known CoMFA method. Once we have the contours, each contour can be encoded by the binary code as illustrated in this chapter. A list of such contours, with specifications of their relative positions, will lead to a 3D code for molecular shapes. Comparisons of 3D shapes will now be more involved. 

When added to CoMFA, the MLP is calculated by Eq. [33] in the same 3D grid as the steric and electrostatic fields. However, the inclusion of MLP in CoMFA is not straightforward because of the composite nature of lipophilicity, which as discussed earlier contains a hydrophobic component and polar terms. In contrast to 2D-QSAR, where the correlation between explanatory variables is easy to assess with a correlation matrix, the correlation between CoMFA signals associated with each molecular field is far from evident. At present, a systematic comparison of all CoMFA models generated by all possible combinations of molecular fields is the only reliable means of detecting correlations between CoMFA results. ... 

---