Date Approved

5-6-2019

Embargo Period

5-7-2019

Document Type

Thesis

Degree Name

MS Computer Science

Department

Computer Science

College

College of Science & Mathematics

First Advisor

Hnatyshyn, Serhiy Y.

Second Advisor

Hnatyshin, Vasil

Third Advisor

Thayasivam, Uma

Subject(s)

Computer algorithms; Chemistry--Analytic

Disciplines

Chemistry | Computer Sciences

Abstract

The huge amount of spectroscopic data in use in metabolomic experiments requires an algorithm that can process the data in an autonomous fashion while providing quality of analysis comparable to manual methods. Scientists need an algorithm that effectively deconvolutes spectroscopic peaks automatically and is resilient to the presence of noise in the data. The algorithm must also provide a simple measure of quality of the deconvolution. The deconvolution algorithm presented in this thesis consists of preprocessing steps, noise removal, peak detection, and function fitting. Both a Fourier Transform and Continuous Wavelet Transform (CWT) method of noise removal were investigated. The performance of the automated algorithm was compared with the manual approach. The tests were conducted using data partitioned into categories based on the amount of noise and peak types. The CWT is shown to be an adequate method for estimating the locations of peaks in chromatographic data. An implementation was provided in Microsoft Visual C# with .NET 5.0.

ThesisCode.zip (93 kB)

Share

COinS