Molecular Orbital Calculations

Many workers have been attracted by the possibility of calculating molecular orbital indexes that should be relevant to the differences in solvation energy between water and octanol. The first such effort was by Rogers and Cammarata in 1969 [38], who developed a correlation equation for octanol/water logP of just 30 simple aromatic solutes using charge density and induced polarization as calculated parameters, suggesting that solvation by the aqueous phase was charge-controlled, whereas solvation by the octanol phase was polarizability controlled. This idea was followed up by Klopman and Iroff [39] in 1981, and by Zavoruev and Bolotin [40] in 1982, who also found charge density calculations useful in simple series of solutes. The Klopman study of 61 solutes indicated that terms for bulk as well as charge were necessary. As many as 10 independent variables were required, including the number of hydrogen, carbon, nitrogen, and oxygen atoms; the sums of squared charges on carbon, nitrogen, and oxygen atoms, and indicator variables for the presence of functionalities such as acid or ester, nitrile, and amide.

Coming now to 1989, Bodor and coworkers [41] developed a regression model for prediction of octanol/water logP in which charge density on nitrogen and oxygen atoms was found important, a dipole moment term was included, and "bulk" was represented by descriptors of surface area, volume, and molecular weight. They extended the Klopman set of solutes to include 118 miscellaneous aliphatic, aromatic and he-teroaromatic compounds, with functional groups embracing alcohols, phenols, amines, ethers, and amides; and even included a few complex drug molecules such as atropine and tetracycline! The most important result to emerge from this work was that surface area (see section 2.3.5) and volume were the most significant descriptors. Although the standard error and correlation indicate a good predictive power, the method, needing the intermediate computation of 15 regression parameters, on a structure previously fully optimized using the AMI procedure, must rank as too unwieldy for general predictive use.

0 0

Post a comment