Corrosion inhibition efficiency of thiophene derivatives on mild steel: a QSAR model

Abstract

A quantitative structure activity relationship (QSAR) model was built using Interval Partial Least Squares and Partial Least Squares (IPLS-PLS) regression for the prediction of corrosion inhibition efficiency of thiophene derivatives. Eleven compounds with their activity expressed as percentage inhibition efficiency (%IE) were obtained and divided into a training set (ntrn = 7) and test set (ntes= 4). Molecular descriptors were generated using Dragon software and the important relevant descriptors were selected using an objective variable selection followed by subjective variable selection using IPLS. Several models were built using PLS regression and the models were evaluated using statistical significance characterization, r2 and root mean square error calibration (RMSEC). The robustness, accuracy and predictive ability of the models were carried out using external and internal cross validation using regression coefficient cross validation (r2 cv) and regression coefficient prediction (r2 pred). The values were calculated and found to be > 0.5 and 0.8 respectively for the first and second model and for the external validation the values are found to be > 0.6 and 0.5 respectively and the r2 value was found to be > 0.9. Application of the built model to calculate the theoretical %IE was obtained and is closer to the %IE experimental. The result showed the predictive ability of the model was good and can be used to design a similar group of compounds with corrosion inhibition efficiency