Molecular indices are characterized by fixed numerical values, which are independent of the physical chemistry property or/and biological activity considered. Once the bonding pattern of a molecule is defined they can then be computed. Contrarily, variable indices depend on the property/activity under consideration. Thus, variable descriptors are functions instead of numerical data and they change during the regression analysis. This more general definition of molecular indices has offered a powerful tool for studies within the realm of QSAR/QSPR theory and it has enabled authors to surmount the so called “nightmare of molecular descriptors”. This drawback arises when one has to choose among a huge number of possibilities to assign a set of molecular descriptors to perform a given study. In fact, the optimization of molecular descriptors for the particular application makes it unnecessary to focus the attention on the choice of a particular set of indices and the efforts can then be directed towards the suitable search of the best mathematical algorithm to obtain the corresponding flexible descriptors.
(Párrafo extraído del texto a modo de resumen)