Projection pursuit based clustering of NMR variables from metabolic response of Hella cells to silica nanoparticles

Exploratory projection pursuit is a set of data analytical techniques for finding “interesting” low-dimensional projections of different multivariate data including NMR [1]. In this work, for the first time, projection pursuit is used for the clustering of NMR data variables. The aim of this work was that which variables or variable regions behave similarly. One of the effective factors in improving the clustering was the change in the size of the bins. Binning is a common technique when dealing with huge NMR data. It is useful for data reduction, solution of peak shift problem, and improvement of sensitivity of quantification [2, 3]. Factors that have been studied to improve clustering efficiency included type of pretreatment and number of factors. However, they showed no effect on the results of the clustering of the variables. Finally the NMR spectra were simply and successfully clustered to different regions without any supervision and additional information. Optimum bin size of about 100 variables per bin was found for obtaining proper information content and clustering results.