QSPR and QSAR Models Derived with CODESSA Multipurpose Statistical Analysis Software

An overview on the development of QSPR/QSAR equations using various descriptor mining techniques and multilinear regression analysis in the framework of program CODESSA (Comprehensive Descriptors for Structural and Statistical Analysis) is given. The description of the methodologies applied in CODESSA is followed by the presentation of the QSAR and QSPR models derived for eighteen molecular activities and properties. The properties cover single molecular species, interactions between different molecular species, properties of surfactants, complex properties and properties of polymers. Historical Introduction The fast progress in modern computer technology has created an entirely new environment for the efficient use of the theoretical constructions of natural science in many areas of applied research. The theoretical approach has proven to be especially beneficial in chemistry and allied sciences, where the experimental study and synthetic development of new compounds and materials can frequently be time consuming, expensive or even hazardous. Contemporary quantum theory of molecular matter and the corresponding ab initio computational methods can, in principle, predict the properties of isolated small molecules with an accuracy comparable to the experimental precision. However, the majority of industrially and environmentally important chemical processes, and all biochemical transformations in living cells take place in heterogeneous condensed media. The extreme complexity of such systems usually prohibits use of ab initio theory and thus the relationship between the chemical and physical properties and the molecular structure in these systems is often poorly described and understood. The direct development of empirical equations that are commonly referred to as the quantitative structureactivity/property relationships (QSAR/QSPR) has been an attractive alternative approach to predict molecular properties in complex systems. Notably, the QSAR methodology has been extremely productive in pharmaceutical chemistry and in computer-assisted drug design. Thousands of potential new therapeutic agents have been first developed on a computer screen before the attempted implementation of selected examples in a synthetic laboratory. In analytical chemistry, QSPR equations are commonly used to predict spectroscopic, chromatographic and other analytical properties of compounds. In recent years, the QSPR approach has been rapidly expanding to diverse areas of industrial and environmental chemistry. In most contemporary applications, empirical molecular descriptors that rely on some experimental data have been used in the development of QSAR/QSPR equations. Such descriptors, ranging from the original Hammett substituent σ-constants to the highly popular partition coefficients between water and octanol (logP) are, strictly speaking, restricted to those compounds for which the necessary experimental data are available. Another shortcoming of experimental descriptors evolves from the fact that many of them reflect a complicated combination of different physical interactions and thus their appearance in a QSAR/QSPR equation may be difficult to interpret. An alternative approach is to use molecular descriptors which can be derived using only the information encoded in the chemical structure of the compound. Importantly, such theoretical descriptors can be developed for compounds that are experimentally unexplored, unavailable, or even unknown. From: AAAI Technical Report SS-99-01. Compilation copyright © 1999, AAAI (www.aaai.org). All rights reserved.