CHEMOMETRICS FOR PLANNED MASS SPECTROMETRY EXPERIMENTS NEAR A COMET

K. Varmuza

Vienna University of Technology, Laboratory for Chemometrics
Institute of Food Chemistry
Getreidemarkt 9/160, A-1060 Vienna, Austria, kvarmuza@email.tuwien.ac.at


For January 2003 the launch of space mission ROSETTA is planned by the European Space Agency. ROSETTA is scheduled to rendezvous with a comet in year 2011 and to enter into an orbit around the comet. One of the instruments on board will be COSIMA, a time-of-flight mass spectrometer for secondary ion mass spectrometry which is dedicated to collect and to analyze cometary grains. Analysis of organic and inorganic components is considered to contribute to the theory about the influence of cometary matter to the development of self-reproducing molecules and finally of life on earth. Prominent tasks within the COSIMA experiment that are closely related to chemometrics are data evaluation, data reduction and data interpretation.

The COSIMA instrument is still in an assembling and testing state, and no cometary material has yet been brought back to earth under controlled conditions. Therefore a set of reference compounds has been measured by an instrument similar to COSIMA. The mass spectra have been transformed into appropriate spectral features (x-variables); the corresponding chemical structures have been characterized by binary substructure descriptors (y-variables). Exploratory data analysis by PLS and PCA showed useful spectra-structure relationships and thereby supported the detection of those structural classes that are well reflected by spectral data. Interpretation of the loading plots indicated spectral properties which are characteristic for particular substance classes. For a couple of substance classes multivariate classifiers have been developed to characterize the chemical composition of a sample.

Responses of multivariate classification and calibration models have been used to characterize a mass spectrum by a vector. This type of mass spectra representation has been shown to be more powerful than the usual representation by peak heights - especially in spectral library searches with unknowns not present in the library.

Collaboration with J. Kissel (principal investigator of the COSIMA project), F.R. Krueger, E.R. Schmid, W. Demuth and W.Werther are gratefully acknowledged. More details about the MS-Chemometrics-COSIMA project are in http://www.lcm.tuwien.ac.at/cosima.