Bioinformatics Solutions Towards the Advancement of Pathogen Detection with Metagenomics

Next-generation-sequencing (NGS) has great potential for use as an excellent tool for detecting and diagnosing infectious disease. As applied to metagenomics, NGS poses several challenges when geared toward general pathogen... [ view full abstract ]

Next-generation-sequencing (NGS) has great potential for use as an excellent tool for detecting and diagnosing infectious disease. As applied to metagenomics, NGS poses several challenges when geared toward general pathogen detection activities. Some of these challenges include the robust assignment of pathogens, given an incomplete database, short reads, and algorithms that focus only on easy use cases (e.g. pathogens comprise most of the sample). General metagenomics taxonomy classifiers have been employed to help identify organisms within clinical samples. However, before NGS can be used as a routine procedure in a clinical setting, several hurdles must be overcome: (1) an easy-to-use environment that technicians or other non-bioinformatics experts can use, including reports and visualizations that can be interpreted by clinicians in a meaningful fashion; (2) rapid bioinformatics tools which run effectively on commodity hardware; (3) levels of confidence for reported organisms that is not tied solely to abundance; (4) a database of known pathogens that allow meaningful reporting.

Here, we provide some examples of the issues surrounding the use of NGS as a method to robustly identify pathogens in complex samples. We also present a series of efforts designed to: (1) lower the barrier for non-experts to use NGS for routine bioinformatics applications by developing a user-friendly web-based suite of tools; (2) limit the number of organisms mis-identified within samples, thereby improving positive predictive value; (3) provide the ability to fine-tune parameters to better assess what defaults should be used given specific questions that require different cutoffs (e.g. pathogen discovery vs detection of known pathogens). We also present a first attempt at developing confidence scoring algorithms that are not tied to abundance of identified organisms. Funding for this project is provided by the Defense Threat Reduction Agency – Joint Science and Technology Office for Chemical and Biological Defense, contract number HDTRA1-15-C-0013.