Why there are AMR gene and species call differences between read and contig data?

Overview

The AMR pipeline looks for AMR sequences within read and contig data (click here for an overview of the pipeline workflow). Identified AMR sequences are then used for pathogen-of-origin detection. You may notice a difference in AMR genes and species detected with contig data versus those identified with read data. Similarly, you may notice a difference between species calls made with the mNGS and AMR pipelines. These discrepancies stem from differences in the pipeline workflow.

AMR Pipeline Read vs Contig Workflows

To identify AMR genes, reads are directly compared against the Comprehensive Antibiotic Resistance Database (CARD) using the KMA mapping method, whereas contigs depend on the assembly pipeline and are compared against CARD using BLAST. Moreover, read matches will only be reported for one out of four detection models used for contig data (see Model definition under AMR Gene Information). These methodological differences may lead to AMR genes detected with read data but not contigs and vice versa.

Regarding species calls, only AMR coding sequences will be used for pathogen-of-origin prediction. Accordingly, AMR sequences that were only detected with read data may lead to the detection of species that were not detected with contig data and vice versa. An advantage of the AMR module is that it takes advantage of both reads and contigs to get a comprehensive list of AMR-associated sequences found in your dataset.