Automatic classification of chromatographic peaks

An intelligent algorithm was developed to automatically categorize chromatographic peaks resulting from the separation of protein mixtures using ion exchange chromatography. A vector quantizing neural network (VQN) was trained and used to classify peaks into six distinct categories based on peak geometry: Gaussian, fronted, tailed, leading shoulder, trailing shoulder, and overlapping. A preprocessing algorithm consisting of noise filtering, vector normalization, and cubic spline interpolation was developed to map peaks to identically sized vectors before introducing them to the VQN. Experimental data was used for training and testing. The VQN correctly classified 90% of the test peaks