Unmixing of Overlapping Peaks in GCxGC-MS by Means of Clustering of Ion Peaks
Saibal Roy
Friday, June 15, 2012, 1:00pm
256C Avery Hall
Committee Members:
Dr. Stephen Reichenbach (Advisor), Dr. Jitender Deogun, and Dr. Ashok Samal
Abstract:
Comprehensive two-dimensional gas chromatography with mass spectrometry (GC×GC-MS) is an advanced technology for separating, analyzing, and identifying compounds in complex chemical samples. Peak detection in GC×GC-MS is the first step in identifying and quantifying compounds present in a sample. Unmixing of overlapping peaks in GC×GC-MS chemical samples is a complex research problem. Researchers have proposed many techniques to unmix overlapping peaks in GC×GC-MS. Most of these approaches are based on complex multivariate mathematics.
This report presents an approach to detect ion peaks and unmix overlapping blobs using clustering in GC×GC-MS data. This ion peak detection algorithm is based on the watershed algorithm used in image segmentation. The unmixing method is based on cluster analysis of ion peaks. The research investigated three algorithms to perform cluster analysis. The research performs ion peak detection and clustering on bio-fuel, diesel, and gasoline samples. Comparisons of the results from different clustering algorithms indicate that DBSCAN and OPTICS unmixes overlapping peaks with higher accuracy than COBWEB. For the experimental data, the clustering method for unmixing is seventy five percent accurate for overlapping peaks with large peak intensities. The accuracy of this method diminishes for chromatograms with narrow blob distribution and overlapping peaks with smaller intensities.