| 초록 |
The n-octanol/water partition coefficient (KOW) is the ratio of the concentration of a chemical in n-octanol to in the water in a two-phase equilibria system. This coefficient is used to understand the uptake, distribution, biotransformation, and elimination of chemicals in the external environment. Although there is an increasing need for these properties, it is often difficult to obtain experimental measurements. So, several works to estimate the log(KOW) with correlation or neural networks from molecular structure; including quantum mechanics; have been developed. However, it can be difficult to obtain the electronic properties used as input parameters as well as reliable estimation result. In this work, we proposed a new group contribution based graph convolution network for the estimation of KOW, which had excellent results, in the estimation for properties of pure organic compounds. Compared to other conventional methods, our methods make significant improvements not only in accuracy at estimation but also in isomer discrimination. Moreover, since only a simple molecular structure is required, instant apply is possible. |
