Laboratory for ChemoMetrics, Vienna (Austria)

MS - Chemometrics - COSIMA

 

Summary of Results:  Chemometrics with PDMS Data

Mass Spectra Transformation

 

Overview

Mass Spectra

Transformation

Exploratory

Data Analysis

Classification of

Substance Classes

Spectral and

Structural Similarity

 

 

 

 

An appropriate mathematical transformation of the original peak list data into a set of suitable features is essential for a successful application of multivariate data analysis.

 

A spectral feature xj is a number that can be automatically computed from a spectrum.

 

Often nonlinear transformations are applied that sometimes consider spectroscopic knowledge. Aim of data transformation is to obtain a set of variables that are better suited for a structure related spectra interpretation than the original peak list data.

 

Some of the spectral features used for the PDMS spectra have already been described for EI mass spectra, others were developed for the particular ion fragmentation reactions occurring in SIMS and PDMS.

 

All features are calculated from mass spectra which were normalized to the base peak intensity. All features are scaled to the range 0 to 100. The total number of features used in this work is 708.

 

Thus each mass spectrum is represented by 708 variables.

 

Spectral features used for multivariate data analysis of PDMS spectra.

feature group

feature description

no of features

1

intensities (% base peak) at single masses (m/z 25-38, 40-150)

125

2

intensities at single masses normalized to local ion current (window of 7 masses, m/z 25-38, 40-150)

125

3

averaged intensities of mass ranges m/z 25-38, 40-70, 71-100, 101-150

4

4

logarithmic intensity ratios, m/z 25-38 and 40-150 with mass differences of 1 and 2

250

5

modulo-14 summation for mass ranges 25-120, 25-250, and 121-250

42

6

autocorrelation for mass differences of 1, 2, and 14-60 in mass ranges 25-120, 25-250, and 100-250

147

7

spectra type features describing the distribution of peaks

3

8

peak series features that characterize the joint presence of peaks at given masses

12

 

 

Other subjects

Overview

Mass Spectra

Transformation

Exploratory

Data Analysis

Classification of

Substance Classes

Spectral and

Structural Similarity

 

 

Start

COSIMA

Top

[ Aims | People | Results | Presentations | Pictures | ROSETTA | COSIMA Instrument | Comet Wirtanen | Literature ]

Info

Last update 2000-12-03