Quantitative structure–property relationship prediction of permeability coefficients for some organic compounds through polyethylene membrane

In this work artificial neural network was constructed and trained for the prediction of the permeability coefficients of various organic compounds through low-density polyethylene, based on quantitative structure–property relationship method. The inputs of this network were theoretically derived molecular descriptors which were selected by the stepwise multiple linear regressions technique. These descriptors are; the number of oxygen atoms in a molecule, area-weighted surface charge of hydrogen bonding donor atoms (HA-dependent HDCA-2), molecular transform index lag 11 weighted by atomic van der Waals volume (Morse-11v), molecular transform index lag 10 weighted by atomic polarizability (Morse-10p) and polarity parameter. In order to assess the accuracy and predictability of the proposed model, the cross- validation and Y-scrambling test were employed. The statistical parameters for cross-validation tests are; R2 = 0.964, PRESS = 0.221. The results obtained showed the ability of developed artificial neural network to prediction of permeability coefficients of various compounds. Also result reveals the superiority of the artificial neural network over the multiple linear regressions model.