Prediction of retention times for a large set of pesticides or toxicants based on support vector machine and the heuristic method

Quantitative structure–retention relationship (QSRR) studies were performed for predicting the retention times (RTs) of 110 kinds of pesticides or toxicants. Chemical descriptors were calculated from the molecular structure of the compounds alone. The QSRR models were built using the heuristic method (HM) and support vector machine (SVM), respectively. The obtained linear model of HM had a square of a correlation coefficient: R2 = 0.913, F = 116.70 with a root mean square error (RMS) error of 0.0387 for the training set, while R2 = 0.907, F = 195.49, and RMS = 0.0408 for the test set. The non-linear model by SVM gave better results: for the training set R2 = 0.966, F = 2420.5, RMS = 0.0231 and for the test set R2 = 0.944, F = 339.7, RMS = 0.0313. The prediction results are in good agreement with the experimental values. And the proposed model could identify and provide some insight into what structural features are related to retention time of these compounds.