Forensic Analysis Modeling and Chemometric Investigation of Molecular Markers from Natural and Engineered Environmental Systems

Abstract

Investigations of natural and engineered environmental systems require achieving a complete characterization and identification of contaminants of concern. However, the differentiation of lipid molecular markers (MMs) that originated from various sources is difficult when based simply on their chemical compositions. A comprehensive fractionation protocol of lipid MMs is thus needed for discriminating contamination sources using chemometric (i.e., mathematical and statistical) techniques. The lack of information about the environmental quality of the coastal environment of Alexandria (Egypt) and the impact of anthropogenic sources of pollution in the area have initiated research to study the different extractable lipid classes and MMs in both natural and engineered environmental systems. Samples representing the complexity of the analyzed Alexandria environment were analyzed qualitatively and quantitatively for their lipid classes. A chemometric approach for the interpretation of the lipid data is presented. This unique approach consists of analysis of variance (ANOVA), discriminant partial least squares (D-PLS), and regularized discriminant analysis (RDA). ANOVA was used to compare the relative magnitudes of sample site and type variances. A variable selection approach based on the PLS regression coefficients was proposed to identify the most important lipid classes with their MMs for the classification, and to improve the results. RDA used a regularized covariance matrix estimate for the conventional statistical discriminant analysis methods. The results in this analysis indicated that: (a) the D-PLS and RDA methods provide satisfactory classification of lipid classes, with RDA being slightly better, and (b) the variable selection strategy was able to improve the classification results, and to help identify the most important lipid classes and contaminant MMs for distinguishing and characterizing the environmental samples.