Improved strategy for comparing microbial assemblage fingerprints.

Abstract

Microbial fingerprinting techniques permit the rapid visualization of entire assemblages in single assays, allowing direct comparison of communities in different samples, where the null hypothesis of such analyses is that all samples are the same. The comparison of fingerprints relies upon the precise estimation of all amplified DNA fragment lengths, which correspond to operational taxonomic units (OTU; analogous, but not equal to, a taxon in macroorganism studies). However, computer interpolation of size standards (and consequently OTU size calling) can be imprecise between gel runs, which can lead to imprecise calculation of similarity indices between multiple assemblages. To account for OTU size calling imprecision, all fragments within a range of sizes (a window) can be combined (i.e., "binned") where the window is as wide as the imprecision of OTU size calling. However, artifacts may occur upon binning samples that may cause samples to appear less similar to each other, caused by splitting of OTU between adjacent bin windows. In this work we present an improved binning technique that accounts for OTU size calling imprecision in the comparison of multiple fingerprints. This technique comprises binning all pairwise comparisons in multiple bin window frames, where the starting size of the window (i.e., frame) is shifted by +1 bp for a total of x frames, where x bp is the width of the maximum bin window size in any binning scheme. Pairwise similarity indices between different community fingerprints are calculated for each of the x frames. To best address the null hypothesis of the community comparison, the maximum similarity value of all x frames is then used in downstream analyses to compare the communities. We believe this binning technique provides the most accurate and least biased comparison between different microbial fingerprints.