engleski

Development of gradient retention model in ion chromatography. Part II: Artificial intelligence QSRR approach

Quantitative Structure-Retention Relationships, QSRR, is a common name for methodology that is predicting chromatographic retention time explicitly on basis of analytes’ molecular- structure. Application of these models can generally short the selection-time for propriate method when dealing with other similar analytes, or can replace the time-consuming “trial and error” approach in method optimization. In this work, artificial intelligence was applied in order to develop good QSRR model. The genetic algorithm was used for selection of the most appropriate molecular descriptors, i.e. descriptors with the highest content of useful information. Artificial neural networks, which are generally known as universal approximators, were taken as QSRR models. The QSRR models were developed for several isocratic elutions, all indicating good prediction ability. The results of QSRR prediction were used for development of isocratic retention model, providing prediction over whole domain of isocratic elutions. In order to enable prediction for gradient elutions, a gradient model based on isocratic data was applied. Although the results indicated slight systematic error, the prediction remained satisfactory good.

QSRR; ion chromatography; sugar analysis; retention prediction; articial intelligence

nije evidentirano