Dalin Yuan

Plasma fatty acid metabolic profiling and biomarkers of type 2 diabetes mellitus based on GC/MS and PLS-LDA

Metabolic profiling has increasingly been used as a probe in disease diagnosis and pharmacological analysis. Herein, plasma fatty acid metabolic profiling including non‐esterified fatty acid (NEFA) and esterified fatty acid (EFA) was investigated using gas chromatography/mass spectrometry (GC/MS) followed by multivariate statistical analysis. Partial least squares‐linear discrimination analysis (PLS‐LDA) model was established and validated to pattern discrimination between type 2 diabetic mellitus (DM‐2) patients and health controls, and to extract novel biomarker information. Furthermore, the PLS‐LDA model visually represented the alterations of NEFA metabolic profiles of diabetic patients with abdominal obesity in the treated process with rosiglitazone. The GC/MS‐PLS‐LDA analysis allowed comprehensive detection of plasma fatty acid, enabling fatty acid metabolic characterization of DM‐2 patients, which included biomarkers different from health controls and dynamic change of NEFA profiles of patients after treated with medicine. This method might be a complement or an alternative to pathogenesis and pharmacodynamics research.

Calibration model transfer for near-infrared spectra based on canonical correlation analysis.

In order to solve the calibration transformation problem in near-infrared (NIR) spectroscopy, a method based on canonical correlation analysis (CCA) for calibration model transfer is developed in this work. Two real NIR data sets were tested. A comparative study between the proposed method and piecewise direct standardization (PDS) was conducted. It is shown that the transfer results obtained with the proposed method based on CCA were better than those obtained by PDS when the subset had sufficient samples.

A modified uncorrelated linear discriminant analysis model coupled with recursive feature elimination for the prediction of bioactivity

To meet the requirements of providing accurate, robust, and interpretable prediction of bioactivity, a modified uncorrelated linear discriminant analysis (M-ULDA) model was developed. In addition, a feature selection method called recursive feature elimination (RFE), originally used for support vector machine (SVM), was introduced and modified to fit the scheme of ULDA. From the evaluation of six pharmaceutical datasets, the M-UDLA coupled with RFE showed better or comparable classification accuracy with respect to other well-studied methods such as SVM and decision trees. The RFE used for ULDA has the advantage of increasing the computational speed and provides useful insights into biochemical mechanisms related to pharmaceutical activity by significantly reducing the number of variables used for the final model.

Alternative moving window factor analysis (AMWFA) for resolution of embedded peaks in complex GC-MS dataset of metabonomics/metabolomics study

Hyphenated instruments, such as GC-MS, have been being widely used in many studies of metabolomics/metabonomics. With the deepening of research, chromatograms become more and more complex and the problem of embedded peaks seems to be ineluctable. In this paper, alternative moving window factor analysis (AMWFA) method is introduced to resolve this problem occurring in metabolomics/metabonomics research. This new method can extract selective information by alternative scanning and comparing between two analytical systems. On the basis of the selective information obtained from chromatograms and spectra of two systems, the AMWFA approach can resolve the embedded peaks in GC-MS responses matrix into pure chromatograms and spectra without any model assumption on the peak shape. The resolution results obtained from one simulated data and two real metabolomics data demonstrate the performance of the proposed approach and indicate that it may be a promising one for analyzing complex data from metabolomics/metabonomics studies.

A New Methodology for Uncovering the Bioactive Fractions in Herbal Medicine Using the Approach of Quantitative Pattern-Activity Relationship

The Quantitative Pattern-Activity Relationship (QPAR) approach has been proposed recently by us and applied to the herbal medicine Radix Puerariae Lobatae and a related synthetic mixture system. Two different types of data from the chromatographic fingerprint and related bioactivity capacities of the samples were correlated quantitatively. The method thus developed provided a model for predicting total bioactivity from the chromatographic fingerprints and features in the chromatographic profiles responsible for the bioactivity. In this work, we propose a new methodology called QPAR-F here, to provide another piece of information: recommending the bioactive regions to facilitate bioassay-guided fractionation and related studies. QPAR-F makes use of chromatographic profiles instead of individual data points utilized in our previous work. The chromatograms of the system concerned are firstly divided into different regions or related fractions representing different groups of constituents. Then different combinations of these regions using the exhaustive searching strategy are processed by the partial least squares (PLS) methods to build models. The optimal models give smaller errors between the predicted and measured total bioactivity capacities. The performance of the proposed QPAR-F methodology is first evaluated by a known mixture system with combinations with active ingredients. The results confirmed that QPAR-F works very well in predicting the total antioxidant bioactivity capacities and the active regions could be correctly identified. These findings are very helpful in planning the bioassay-guided fractionation. For this data-mining process, only limited chemical and bioactivity information of the original samples or crude extracts are required. No prior knowledge of activities of the fractions under study is needed. The QPAR-F methodology was also applied to the herbal medicine, Radix Puerariae Lobatae and similar predicted models give smaller errors between the predicted and measured total antioxidant bioactivity capacities could be successfully built.