Methods development and applications of chemometrics techniques in chemical and biochemical studies

Abstract

Owing to the rapid development of instrumentation in analytical chemistry, new and efficient data analysis techniques are required for data interpretation. Chemometrics is a new discipline in chemistry that applied mathematical, statistical and other logic-based methods to analyze chemical data in particular in analytical chemistry. From chemical literature, chemometrics techniques have been applied successfully in different areas. The aim of this project is to apply chemometrics techniques such as wavelet transform and factor analysis to analyze data from chemical and biochemical system. Recently, wavelet transform (WT) was applied successfully in chemistry for data compression and denoising. In this project, WT was employed to process and enhance analytical data in two major areas, they were infrared (IR) spectroscopy and high performance liquid chromatography coupled with a diode array detector (HPLC-DAD). WT was selected as a new compression method to reduce the size of a small IR spectral library. The wavelet compressed library was employed for spectral library searching. The performance of WT in data compression and library searching were compared with fast Fourier transform. A data length of 2p with p being any integer is usually required for wavelet computation. A new method called coefficient position retaining (CPR) method was developed in this work to handle experimental data with a length not equal to 2p. Derivative spectroscopy is another signal processing technique that commonly used in analytical chemistry for data analysis owing to its popularity on the apparent higher resolution of the differential data when compared with the original data. Derivative of an analytical signal is usually derived from numerical differentiation. However, this method has a major drawback in increasing the noise level in computing higher order derivatives. In order to solve this problem, WT was proposed as a new method for approximate derivative calculation. Results indicated that the proposed method is much better than the conventional numerical differentiation method. The development of hyphenated instrument such as HPLC-DAD leads to the development of new type chemometrics technique for data interpretation. Heuristic evolving latent projections (HELP) algorithm is one of the most popular method for analyzing data from HPLC-DAD. This method can extract most of useful information from the raw experimental data. FWT was proposed as a pre-processing step of the HELP algorithm. Both compression and denoising properties of FWT could enhance the HELP algorithm especially for HPLC-DAD data with low signal-to-noise ratio (SNR). This new algorithm (FWT-HELP) was identified as a potential tool for analyzing chemical constituents in traditional Chinese medicine (TCM). In terms of chemistry, chemical constituents of Chinese herbs are a complex black system. There is no a priori information concerning the chemical composition of the samples. Both FWT and HELP algorithms have its own advantages which can simply and enhance the chemical analysis of TCM. A traditional Chinese medicinal herb, Cordyceps sinensis (冬蟲夏草), was selected as an example to test the performance of the proposed algorithm.