Calculation of Octanol/Water Partition Coefficients (logP) using Artificial Neural Networks and Connection Matrices
Klaus‐Jürgen Schaper, Maria Luisa Rosado Samitier · Quantitative Structure-Activity Relationships · 1997
Abstract Artificial neural networks of the backpropagation type with three layers (input, hidden, output) are able to recognize structural molecular features determining the lipophilicity of unionized organic molecules and to directly use the chemical structure for the calculation of the partition coefficient [logP(octanol/water)]. Using a network with three neurons in the hidden layer and binary variables to indicate the presence or absence of atom types and bond types a standard deviation of s = 0.248 was obtained in the correlation between logPobs. and logPcalc. for a small training set of 268 simple organic molecules containing C, H, N, O, S, halogens. For a test set of 50 similar molecules the predicted logP values were satisfactory (s = 0.659).