Logit-Based Neutral Zone Classification Algorithm with Application to Microbial Community Profiling

• Activity Number: 524
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Rebecca Le*+ and Daniel R. Jeske
• Companies: University of California, Riverside and University of California, Riverside
• Keywords: classification ; neutral zone classifier ; generalized linear mixed model ; microorganism application

Abstract

Microbial community compositions are identified in a cost-effective manner by a method called Oligonucleotide Fingerprinting of rRNA Genes (OFRG). Hybridization signal intensities between gene clones and probes are observed and classified into classes of non-binding, partial binding or complete binding to build OFRG fingerprints which are then clustered into microbial communities. Neutral Zone Classifiers (NZC) have been recently created to improve the fidelity of this approach by assigning class labels N for any ambiguous hybridization values under the prior assumptions of independent data among probes and Gaussian mixture class distributions. To relax these assumptions, a Logit-based Neutral Zone Classifier (LNZC) is developed using a Generalized Linear Mixed Model. Findings are discussed and compared with the existing NZC algorithm.