Using artificial neural network tools to analyze microbial biomarker data
A major challenge in the successful implementation of bioremediation is understanding the structure of the indigenous microbial community and how this structure is affected by environmental conditions. Culture-independent approaches that use biomolecular markers have become the key to comparative microbial community analysis. However, the analysis of biomarkers from environmental samples typically generates a large number of measurements. The large number and complex nonlinear relationships among these measurements makes conventional linear statistical analysis of the data difficult. New data analysis tools are needed to help understand these data. We adapted artificial neural network (ANN) tools for relating changes in microbial biomarkers to geochemistry. ANNs are nonlinear pattern recognition methods that can learn from experience to improve their performance. We have successfully applied these techniques to the analysis of membrane lipids and nucleic acid biomarker data from both laboratory and field studies. Although ANNs typically outperform linear data analysis techniques, the user must be aware of several considerations and issues to ensure that analysis results are not misleading: (1) Overfitting, especially in small sample size data sets; (2) Model selection; (3) Interpretation of analysis results; and (4) Availability of tools (code). This poster summarizes approaches for addressing each of these issues. The objectives are: (1) Develop new nonlinear data analysis tools for relating microbial biomolecular markers to geochemical conditions; (2) Apply these nonlinear tools to field and laboratory studies relevant to the NABIR Program; and (3) Provide these tools and guidance in their use to other researchers.
- Research Organization:
- Oak Ridge National Lab. (ORNL), Oak Ridge, TN (United States); University of Tennessee, Knoxville, TN
- Sponsoring Organization:
- USDOE Office of Science (SC)
- OSTI ID:
- 895288
- Report Number(s):
- CONF-NABIR2004-20; TRN: US200702%%740
- Resource Relation:
- Conference: Annual NABIR PI Meeting, March 15-17, 2004, Warrenton, VA
- Country of Publication:
- United States
- Language:
- English
Similar Records
Waste Characteristics of the Former S-3 Ponds and Outline of Uranium Chemistry Relevant to NABIR Field Research Center Studies
Application of Nonlinear Analysis Methods for Identifying Relationships Between Microbial Community Structure and Groundwater Geochemistry