The Criticism That Is Not

A few years ago, Dearden and collaborators [14] published a paper with the very provocative and pretentious title How Not to Develop a Quantitative Structure-Activity or Structure-Property Relationship (QSAR/QSPR). In this publication, in the section Errors in QSARs/QSPRs, the authors listed, in total, 21 types of error in QSAR/QSPR development and use They ended their list with Lack of mechanistic interpretation as an error, which itself is not an error but a matter of politics  

as the acronym implies, is about quantitative structure-activity relationships and not about quantitative mechanistic-activity relationships We will comment on few of the errors on the list of 21 types of error, which, in our view, are not errors and on a few questions for which have been offered answers for those interested in considering them. It is not our aim here to critically review the paper How Not to Develop a Quantitative Structure-Activity or Structure-Property Relationship by Dearden et al but to make our contribution to How to Develop a Quantitative Structure-Activity or Structure-Property Relationship. One particular matter that we will address is, in our view, a misrepresentation in the article by Dearden et al. of the connectivity index and the variable connectivity index. 

As mentioned, the problem has been solved, and it was solved almost 25 years ago [7] But to know about it, one should read QSAR publications coming from mathematical chemistry circles. Obviously, lack of cross-communication between the Hansch QSAR school and the circle around chemical graph theory is the culprit with equal guilt on both sides. 

The issue of collinearity should not be raised to the level of doctrine, dogmatism, and politics. It is a matter of mathematics. That one so important result can, for so long, be overlooked by a sizable group of researchers in QSAR, if not the majority, is hardly just a matter of unfamiliarity with the literature on the topic by so many people. 

As to their reasons for undesirable use of collinear descriptors, Dearden et al. state that (i) two highly collinear descriptors are effectively contributing the same 

---