For the development of new fungicides against rice blast, the quantitative structural–activity relationship (QSAR) analyses for fungicidal activities of thiazoline derivatives were carried out using multiple linear regression (MLR) and neural network (NN). We have studied the substituent effects at
para site of R
1 and at three sites (
ortho,
meta, or
para) of R
2 aromatic rings in compounds. The results of MLR and NN analyses in the training set of Set-3 showed good correlations (
r2 values of 0.829 and 0.966, respectively) between the descriptors and the fungicidal activities. Five descriptors including the non-overlap steric volume
bacaa460b10f80b4476e2c23205195a"" title=""Click to view the MathML source"" alt=""Click to view the MathML source"">(SVR2C2), Connolly surface area
(SAR1), hydrophobicity
(∑πR2), and Hammett substituent constants (
σpR1 and
σmR2) were selected as important factors of fungicidal activities. Although the descriptors of optimum MLR model were used in NN, the results were improved by NN. This means that the descriptors used in MLR model include non-linear relationships.