|
Activity Number:
|
285
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #310253 |
|
Title:
|
Methods for Classification of NMR Metobololomic Spectroscopy
|
|
Author(s):
|
Yuanxin Xi*+ and David Rocke
|
|
Companies:
|
University of California, Davis and University of California, Davis
|
|
Address:
|
Institute of Data Analysis and Visualization, Davis, CA, 95616,
|
|
Keywords:
|
NMR ; Adaptive Binning ; Classification ; Biomarker Discovery
|
|
Abstract:
|
We present a classification case of 2D-NMR TOCSY datasets for biomarker discovery in the study of Pseudoxanthoma elasticum (PXE), a heritable recessive connective tissue disorder. PXE is primarily caused by metabolic disorder and we use in-vivo 2D-TOCSY spectroscopy to detect small molecular metabolites for 21 patients and 19 controls. The datasets are pre-processed by an adaptive binning based on peak probability densities and then compared by logistic regression and support vector machines. The results show this adaptive binning method is more stable than classical fixed size binning. The two classes are well separated in PLS component spaces and the loading plot indicates the significant regions that suggest potential biomarkers.
|
