Modeling Changes in Bacterial Prevalence with Sequenced Data

• Activity Number: 421
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Raymond G. Hoffmann*+ and Ke Yan and Jessica VandeWalle and Sandra McLellan
• Companies: Medical College of Wisconsin and Medical College of Wisconsin and University of Wisconsin at Milwaukee and University of Wisconsin at Milwaukee
• Keywords: wavelet analysis ; harmonic regression ; time series ; normalization

Abstract

Information about the prevalence of bacterial taxa in water samples can be determined via next generation sequencing. Samples were taken from two different, but related, sites in lake Michigan over a three year period. Since the samples were taken during different seasons and at different points in time, determining temporal effects required methods that were both flexible and did not required regular measurements. Initially there were over 600 taxa; after aggregation of similar taxa there were still 22 time series with day specific and site specific co-variate information. Wavelet analysis after filtering to reduce sampling noise was used to determine the temporal aspects of the multivariate time series because of the irregular occurrence of the sampling points over time. Initial attempts with sine-cosine time series failed because of the irregularity of the time points and shapes of the prevalence curves. Since the data was obtained as separate samples, different methods of normalization were also explored. The main goal was to determine if wavelet coefficients could be used to identify common patterns of temporal response among the major taxa and sub-taxa.