Date of Award

9-20-2010

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Bioinformatics

First Advisor

Robert Shmookler Reis

Abstract

Although gas chromatograpy-mass spectrometry (GC-MS) is a mature technology for profiling primary metabolites, analyzing those datasets in a large-scale and high-throughput manner is a great challenge because of the complex nature of the data. There is often non-linear inter-run variance in the retention time between runs, making it difficult to identify corresponding peaks among multiple runs. Moreover, peaks are rarely homogeneous. They often contain systematic signals, i.e., background, or non-systematic signals, i.e., noise, and many peaks comprise multiple co-eluting or overlapping components. Thus, identifying all the pure components in all of the peaks from the entire dataset, and comparing them efficiently between runs, is a great challenge and rate-limiting step in the data processing part of the work flow of GC-MS analysis. Parallel factor analysis (PARAFAC) is a simple but very powerful multi-way method, but it has been used in a very limited way for global metabolite profiling. In this study, we successfully applied PARAFAC to our GC-MS data, in order to find putative biomarkers for longevity in C. elegans. A systematic method for the application of the model to entire datasets was developed and implemented, in which the model's prerequisite for data trilinearity was satisfied by using a novel alignment method, an iterative-block shifting approach.

Download

Share

COinS