Optimizing Oligonucleotide Fingerprint Classification for Microbial Community Analysis

• Activity Number: 250
• Type: Contributed
• Date/Time: Tuesday, August 5, 2003 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): James Borneman and S. James Press*+ and Katechan Jampachaisri and Lea Valinsky
• Companies: University of California and University of California, Riverside and University of California, Riverside and University of California, Riverside
• Keywords: genomics ; classification ; clustering ; Bayesian ; trees ; hierarchical

Abstract

The study of microbial community composition through sequence analysis of rDNA clone libraries is often impractical because of the high costs associated with examining such diverse communities. Oligonucleotide fingerprinting of ribosomal RNA genes (OFRG) is an alternative, array-based approach that sorts rDNA clones into taxonomic groups through a series of hybridization experiments. For every hybridization experiment, the signal intensities are transformed into three discrete values 0, 1, and n, where 0 and 1 respectively specify negative and positive hybridization events and n designates an uncertain assignment. Various approaches were tried to resolve the uncertainty. These include Bayesian classification with the normal distribution, Bayesian classification with the exponential distribution and gamma distribution, along with tree-based classification. All 0-1 data produced from each classification approach, including the original 0-1-n data, were afterwards clustered using UPGMA. The performance of each approach was compared with results in known 0-1 reference data. The comparisons indicated that the approach using Bayesian classification with normal densities was best.