Document Type

Dissertation - University Access Only

Award Date

2004

Degree Name

Doctor of Philosophy (PhD)

Department / School

Nutrition, Food Science, and Hospitality

First Advisor

Chunyang Wang

Abstract

The accurate analysis of isoflavones and their metabolites is extremely important in understanding their bioavailabilities and biochemical functions. Preparation of urine isoflavone analysis can be time consuming. Three methods, including a method developed in our laboratory, were compared and discussed. The method that we developed had significantly higher recoveries and required much less time. A three-day, multiple freeze[1]thaw study was conducted on urine isoflavones. Results from the multiple freeze-thaw study showed a significant decrease on day 3 for daidzein (DAI) (P<0.0002).

UV-Visible spectrophotometry was used in the development of a rapid method in determining isoflavone concentrations in urine samples. Various sample treatments were completed before the scanning including: original urine samples, urine sample dilution, urine solid phase extraction, urine liquid-liquid extraction, and liquid-liquid spiked extraction of spiked samples. The samples were scanned and mathematical models were developed to predict isoflavone concentrations based on spectrophotometric properties. A liquid extraction was found to be the most effective method yielding R 2 of 74 and 89% for DAI and GEN, respectively. T-values were found to be insignificant for DAI (-0.15) and slightly significant for GEN (2.06). Samples collected from a dietary intervention study were used as unknowns. The unknown samples yielded R 2 values of 34 and 53% and non-significant t[1]values of 1.40 and 0.66, for DAI and GEN, respectively. UV-Visible spectrophotometry is simple, cost-effective and accurate and could be an enormous contribution in soy health studies.

Neural networks have been implemented in many applications from toys to biological neural networks. Urine samples were collected from a dietary intervention study and treated using solid phase extraction. Isoflavone extracts were then analyzed by HPLC-UV and scanned using UV-Vis spectrophotometry. Mathematical models were developed to predict isoflavone concentrations based on spectrophotometric properties using SAS. The sample set was then used to train a neural network to predict the isoflavone concentration. Isoflavone concentrations from the predicted models and neural networks were evaluated in comparison to the HPLC reference method. The mathematical models developed from SAS yielded R 2 of 10 and 26% for DAI and GEN, in the validation set, respectively. T-values were found to be significant for DAI (2.43) and for GEN (5.31). These models applied to the unknown sample set, yielded R 2 values of 6 and 12% and significant t-values of 2.35 and 3.96 for DAI and GEN, respectively. The neural network yielded R 2 of 84 and 83% for DAI and GEN, respectively. T-values were found to be non-significant for DAI (-1.04) and for GEN (-0.73). The unknown samples yielded R 2 values of 72 and 75% and non-significant t-values of -0.84 and 0.55, for DAI and GEN, respectively.

Library of Congress Subject Headings

Isoflavones -- Analysis.
Soyfoods -- Therapeutic use.

Publisher

South Dakota State University

Download

Share

COinS
 

Rights Statement

In Copyright