Desiderata for Molecular Descriptors

Should have good correlation with at least one property 

Should change gradually with gradual change in structures 

Should have the correct size dependence, if related to the molecule size 

It should be mentioned here that the Occam s razor rule has been mentioned from time to time by those who continue to object to the apparent explosion of topological indices, which has occurred in more recent years in QSAR, but the simplest of competing theories being preferred does not extend to replacing theories with descriptors. Therefore, the simplest of competing descriptors should not necessarily be preferred. What is important for molecular descriptors is not how similar they are and how much they may duplicate one another, but do they differ one from another in some details that are important (for the structure-property-activity relationship). 

Clearly cases (i-iii) are not broad enough to include molecular properties as constituents of a structure, but with some imagination (and good will) perhaps case (iv) can even if properties are not a synonym for structures. 

---