Summary of Results: Chemometrics with PDMS Data Overview |
|
On board of COSIMA a pre-evaluation of mass spectrometric data is essential for several
reasons:
(a)
A selection and reduction of data is necessary before
transmission to ground.
(b)
A time gap of several days must be expected between checking
COSIMA data on Earth and sending new commands.
Consequently, the COSIMA
instrument has to work autonomously to some degree. Decisions
have to be made on board for instance like "which new experiments should
be made ?". Such decisions should be mainly based on chemical structure
information derived from data measured on cometary matter.
The expected complexity of
cometary material will not allow substance identifications but has to focus on the
recognition of substance classes or functional groups. A realistic aim of on
board data evaluation is a chemical structure oriented characterization of the
analyzed samples or sites.
Chemometric methods were
already successfully applied for the automatic recognition of substructures or other structural properties from low
resolution electron impact (EI) mass spectra. In some cases even a systematic structure elucidation is possible. From the given
molecular formula and restrictions about the presence or absence of
substructures (automatically obtained from spectra) an exhaustive set of
possible chemical structures can be constructed by an isomer generator
software.
The methods of multivariate
data analysis applied to spectra interpretation are all based on the
characterization of spectra by a set of variables (spectral features). A spectrum then can be considered as a point in a
multidimensional space with the coordinates defined by these spectral features.
Several mathematical procedures are available to "look" into the high
dimensional space (exploratory data analysis, cluster analysis) or to find
decision rules (classifiers) capable to separate for instance a substance class
from all other compounds.
Some of these chemometric
methods have been tested for the PDMS data. The main difference to the
treatment of EI mass spectra is a somewhat different transformation of the
spectra due to other fragmentation processes in PDMS and SIMS.
The application of multivariate data analysis to spectra interpretation typically requires the
following steps:
(a)
building a spectra set,
(b)
transformation of the peak list data into a set of spectral features,
(c)
selection of the most relevant features,
(d)
application of principal component analysis (PCA) or multivariate
classification or calibration methods.
The same strategy has been
applied to the PDMS data considering the following COSIMA related aims:
u development of classifiers for the recognition of substance
classes,
u development of calibration models for a semi-quantitative prediction of
numerical structural properties,
u development of a spectra similarity criterion that reflects structural
similarity.
Other
subjects
|
[ Aims | People | Results | Presentations | Pictures | ROSETTA | COSIMA Instrument
| Comet
Wirtanen | Literature
] |
Last update 2000-12-03 |