Mapping Changes in Soil Microbial Community Composition Signaling Bioremediation

Publisher Website
Google Scholar
Abstract
Chemical signatures of biological processes reflect their complex interrelationships. The chemical profile is rich in information but poor in content due to the complex processes underlying the chemical composition of natural biological communities. A nonlinear mapping technique, based on artificial neural networks (ANNs), was proposed to highlight information coded in lipid signatures in soil by demonstrating the biological response to hydrocarbon contamination. ANNs do not require mechanistic assumptions, and they can cope with nonlinear associations. Soil sample lipid signatures were mapped using ANNs to recover information on exposure to contamination, to assess the potential for bioremediation as assessed by polymerase chain reaction (PCR)/deoxyribonucleic acid (DNA) gene probes, and to monitor the effects of selected inocula. A two-coordinate system was built from signature lipid biomarkers containing 64 components from which the values of target parameters (6 components) could be recovered. The map tracks bioremediation, as characterized by the target parameters, and provides information on how parameters interreact during bioremediation. Using 23 soil sample signatures, a map was built from which the 6 target parameters could be recovered with 4.7% average error. Principal component analyses and nonlinear factor analyses by autoassociative ANN were compared to the nonlinear mapping information. Although these methods provided a good description of signature shift, they did not discriminate among all target parameters.