O588 2-hour Oral Session Resistance mechanisms in Gram-negatives

Effective use of resistance databases for whole and metagenomic sequencing data

Anupam Das*1, Xavier Basil Britto1, Guy Cochrane2, Samir Kumar-Singh1, Frank M. Aarestrup3, Herman Goossens1, Surbhi Malhotra-Kumar4

1University of Antwerp, Laboratory of Medical Microbiology, Wilrijk, Belgium

2Ebi Embl, Cambridge, United Kingdom

3Dtu, Copenhagen, Denmark

4University of Antwerp, Laboratory of Medical Microbiology, Wilrijk, Antwerpen, Belgium

Background: Unrestricted use of has resulted in rapid emergence of antibiotic resistant pathogens, leading to a public health crisis. This problem is further compounded due to the increasing acquisition of resistance and evolution of pathogens to multidrug resistant . Mining of information on antibiotic resistance genes is available from specialized databases referred to as antibiotic resistance gene databases. The databases, however, might limit the scope of researchers to effectively mine information from them, due to lack of regular updates or inconsistent data. We tested three of the most popular databases namely, Antibiotic Resistance Gene Database (ARDB), Comprehensive Antibiotic Resistance Gene Database (CARD) and Resfinder for their ability to accurately predict resistance genes.

Material/methods: We performed an assessment of the available databases with whole genome shotgun metagenomic sequencing data from our in-house collection of clinical samples and from publicly available data. From the publicly available data, we randomly selected a set of 10 genes that include 5 blaVIM and 5 blaNDM genes and their variants. These sequences were submitted as query sequences using the Blast tools available on the three databases using the default parameters.

Results: Results indicated that the CARD and Resfinder databases predicted all the genes accurately using the blaVIM and blaNDM genes as query, while the ARDB predicted 3 of the query genes correctly, 2 predictions were incorrect and the rest of them were not found (Figure(a)). Utilizing the metagenomic sequencing data, CARD database predicted the maximum number of resistance genes – a total of 11, including 2 aminoglycoside resistance genes, 7 β-lactamases and 2 either undefined or other. While the Resfinder detected a total of 2 including 1 aminoglycoside and 1 β -lactamase, the ARDB predicted a total of 4, 1 each from aminoglycoside, β -lactamase, tetracycline and others (Figure(b)).

Conclusions: Based on our results, we suggest that CARD and Resfinder are ideal while using gene sequences as query. However, using metagenomic data, CARD performs better than the rest. The Resfinder database, which was found up to date and accurate, currently detects only acquired genes and ignores chromosomal mutations. The ARDB is limited in its scope due to lack of regular updates.

Figure: Comparison of the 3 databases based on Blast results. Results obtained using blaNDM and blaVIM genes as query (a). The figure key is color coded based on accurate predictions, innacurate predictions and missing records; Whole genome shotgun metagenomic sequences used as query (b). The figure key shows the detected genes categorized as aminoglycoside resistance, β-lactamases, Tetracycline resistance, undefined and other genes.