Abstract
We have been constructing a database named Transition State Data Base (TSDB) in order to develop synthesis routes in silico. The ability to predict experimental yields for synthesis reactions is very important for determining which synthetic route is better than others. For this purpose, we have been investigating the fusion of computational and information chemistry. In the present study, we tried to predict yields using the statistical method with explanatory variables as reaction and activation energies, calculated from molecular orbital (MO) calculations. The analysis also used experimental parameters such as dielectric constants of solvents and reaction temperatures. The present study adopted these parameters and applied the GA-PLS method to aromatic nucleophilic substitution (SNAr) reactions producing a 1,4-dihydro- quinoline-3-carboxylic acid skeleton. It was confirmed that the GA-PLS analysis adopting the calculated values as well as the experimental conditions could expect the experimental yields. However, we have to divide experimental data into 3 categories depending on the isolated techniques such as crystallization, re-crystallization and column
