One of the main sequence analysis tasks we want to perform on the UEGP dataset is the evaluation of antibiotic resistance potential in the wastewater microbial community. We have examples of analytical approaches to this problem in several prior studies.

• A very straightforward approach would be to use the CARD database and associated RGI tools to identify components of the “resistome”. However, RGI works on translated protein sequences, ignoring very short sequences, is optimized for genomes and genomic contigs, and has a 20MB limit if used online. CARD has been used and cited in a few metagenomics studies, but not primarily by direct analysis of shotgun metagenome reads. Instead, the database has been downloaded and used as a target for sequence alignment with an aligner optimized for large NGS problems, such as BWA, as was done in Noyes et al 2016.
• Port et al (2014) published a description of a pipeline (for which software does not yet appear to be available) based on MG-RAST. Communities were scored as to gene transfer potential — defined by the presence of plasmids, transposable elements, and phages, antibiotic resistance gene potential — defined by the presence of ARGs, and pathogenicity potential — defined by the presence of known pathogenic bacteria and virulence genes.  I like the approach of splitting the potential for pathogenicity into these three components as all would be required for a scenario in which human-created environments saturated with ambient antibiotics encourage the spread of resistance.
• Yang et al (2016) have developed an integrated database of resistance genes (with information collected from both CARD and ARDB) and tools for searching. Software is available: ARGs-OAP. They remove uncategorized and hypothetical sequences from the reference databases. The analysis is run on a Galaxy server with a local prescreen. This is really ARGs-only and does not include the aspects of transfer potential and pathogenicity potential considered in the previous method.
• Nesme et al (2014) performed a large-scale meta-analysis of multiple metagenomes using a BLAST best hits based strategy with threshholds appropriate for short reads, against antibiotic resistance elements in the ARDB.

Strategies described here are examples but do not fully explore the literature; I continue to delve.