Tampa

Abstract

Metagenomic taxonomic profiling aims to predict the identity and relative abundance of taxa in a given whole genome sequencing metagenomic sample. A recent surge in computational methods that aim to accurately estimate taxonomic profiles, called taxonomic profilers, have motivated community driven efforts to create standardized benchmarking datasets and platforms, standardized taxonomic profile formats, as well as a benchmarking platform to assess tool performance. While this standardization is essential, there is currently a lack of tools to visualize the standardized output of the many existing taxonomic profilers, and benchmarking studies rely on a single value metrics to compare performance of tools and compare to benchmarking datasets. Here we report the development of TAMPA (Taxonomic metagenome profiling evaluation) , a robust and easy-to-use method that allows scientists to easily interpret and interact with taxonomic profiles produced by the many different taxonomic profiler methods beyond the standard metrics used by the scientific community. We demonstrate the unique ability of TAMPA to provide important biological insights into the taxonomic differences between samples otherwise missed by commonly utilized metrics. Additionally, we show that TAMPA can help visualize the output of taxonomic profilers, enabling biologists to effectively choose the most appropriate profiling method to use on their metagenomics data. TAMPA is available on GitHub, Bioconda and Galaxy Toolshed at https://github.com/dkoslicki/TAMPA, and is released under the MIT license. Introduction Metagenomics has become an essential tool to study microbiomes due to improvements in technology and bioinformatic algorithms. Taxonomic metagenome profiling aims to predict the identity and relative abundances of taxa in a given whole genome sequencing (WGS) metagenomic sample. A recent surge in computational methods that aim to accomplish this, called taxonomic profilers, have motivated community-driven efforts to create standardized benchmarking datasets , standardized taxonomic profile formats , as well as a benchmarking platform to assess tool performance on simulated data . While this standardization is essential, there is currently a lack of tools to visualize the standardized output of the many existing taxonomic profilers, and benchmarking studies rely on a single value metrics to compare performance of tools and compare to benchmarking datasets. Indeed, the only two such WGS taxonomic profiling visualization and analysis tools that do exist are either integrated into a single taxonomic profiling method , or else lack the flexibility and interpretability for the analysis and visualization of multiple taxonomic profiles . Neither of these methods are designed for or compatible with the community-driven output formats previously mentioned. Despite the availability of flexible and interactive visualization tools in the area of amplicon microbial analysis (such as 16S rRNA studies), similar methods are yet to be developed for WGS metagenomics. For example, metacoder 8 is a tool that allows for visualizing, analyzing, and manipulating amplicon microbial data. However, metacoder is not designed for WGS metagenomic analyses and cannot be used for analysis and visualizing metagenomic taxonomic profiles due to amplicon analyses relying on Operational Taxonomic Units, a concept that is not relevant to metagenomic studies. Similarly, the recently published preprint for the software package EMPress is an interactive phylogenetic tree viewer not explicitly intended for the visualization of WGS taxonomic profiles. Additionally, lack of tools that provide an interpretable visualization of multiple taxonomic profiles limits the ability of the biomedical community to select a tool. As such, when WGS metagenomic data is generated and a scientist wishes to determine which of the dozens 10–23 of taxonomic profilers to use, they typically rely on benchmark studies 1,24,25 . These benchmark studies often use simulated data that does not accurately reflect their samples of interest. Alternatively, they can run their own simulation and benchmarking study tailor to their use-case, but this requires significant time investment . Often scientists resort to simply picking a familiar tool regardless of its performance characteristics. Given the substantial variability in the performance of taxonomic profiling tools , this may result in misinterpretation of their data and can potentially lead to unfortunate situations where utilizing a single taxonomic profiling tool can lead to an interpretation of data 26 (i.e. presence of Bubonic plague in the New York subway system) that is later to be found to be inaccurate . Furthermore, after a taxonomic profiler is utilized, scientists can encounter difficulty when interpreting statistical measures of differences between the estimated taxonomic frequencies and the ground truth, as well as when comparing differences between tools. This difficulty is further compounded when new or emerging statistical measures are used to characterize differences between metagenomic samples, where confusion can arise about how to interpret such measures 9, 28. To empower biomedical researchers with a robust and easy to use metagenomic taxonomic profile analysis and visualization platform, we have developed a software package TAMPA (Taxonomic metagenome profiling evaluation). Our platform assists scientists in contextualizing, assessing, and extracting insight from taxonomic profiles produced by multiple taxonomic profilers when applied to either real or simulated data. TAMPA is designed to allow users to effectively analyze one or more taxonomic profiles produced by any of the numerous taxonomic profiling methods. Additionally, TAMPA can operate on the widely utilized and community developed BIOM 29 and CAMI 1 profiling formats. We demonstrate the utility of TAMPA by showing how it illuminates the important biological differences between samples and conditions otherwise missed by commonly utilized statistical metrics. When gold standard taxonomic profiles are available, we show how TAMPA can augment existing benchmarking platforms such as OPAL by being incorporated within the tool and providing an interpretable visualization of the profiles . Additionally, we show that TAMPA can enable biologists to choose an appropriate profiling method to use on their real data when a ground truth taxonomic profile is not available, since TAMPA allows users to quickly ascertain similarities or differences in predictions made by multiple taxonomic tools.