Prediction of the retention in thin layer chromatography screening systems by atomic contributions Original Research Article

A novel atomic-contribution system for predicting of image values is presented and validated on 13 thin layer chromatography screening systems on silica gel, where the large experimental datasets (198–761 image values) are available. The image is predicted with error less than 0.5 in majority of solutes (besides several outliers), which corresponds to difference in image equal to 0.28 in the worst case. The system was validated by dividing the data into training and validation datasets, proving its accuracy. The main reason of larger errors in outliers are: large conjugated heterocycles, quarternary ammonium cations, large amount of polar atoms or very simple but unique molecules. The calculations are very easy and can be performed on free software or even manually. The presented method can be used in the retention prediction of new solutes in existing chromatographic screening systems.