Statistical and Computational Methods for Accurate Characterization of Microbes in Clinical and Environmental Sequencing Samples

• Activity Number: 393
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Solaiappan Manimaran*+ and W. Evan Johnson
• Companies: Boston University and Boston University School of Medicine
• Keywords: Clinical Metagenomics ; Pathogen detection ; Next-generation sequencing ; Bayesian modeling ; Software ; Disease outbreak control

Abstract

The rapid identification and quantification of pathogens present in a clinical sample is of high importance in controlling contagious diseases during an outbreak. For example, during the European E. coli outbreak of 2011, there was a 3 week delay in the correct identification of the pathogen strain O104:H4 which caused 3,800 infections and 54 deaths. Here, we present a Bayesian statistical methodology based on a penalized mixture modeling approach that accurately identifies the pathogens with a confidence region to provide accurate diagnosis and best possible treatment. We also present a novel methodology to normalize metagenomic samples and conduct tests of differential expression between two samples. We collected samples containing the hemorrhagic E.coli O104:H4 and performed a simulation with varying levels of read coverage. Our method was able to accurately identify and quantify the pathogen strain even with very low read coverage. We also performed a mixture simulation that contains varying levels of closely related strains and our method was still able to accurately identify and quantify the pathogen strain even when the mixture sample contains closely related strains.