RGI version 6.0.0 – Major release, with parameter and reference data loading changes that may impact scripts. Please see updated README. Added an option (–keep) to keep Prodigal CDS when used with –clean. Removed option –exclude_nudge and replaced with –include_nudge for RGI main so nudged results no longer included in default RGI main output. Added options to use all CARD detection models within RGI bwt, by default RGI bwt now uses protein homolog models only. Added dask and replaced some os.system calls to improve performance. Consensus sequences are reported at depth of at least 5 when using RGI bwt with KMA. Updated RGI load and RGI auto_load commands.
In earlier versions of RGI, by default all Loose matches of 95% identity or better were automatically listed as Strict, regardless of alignment length. At that time, this behaviour could only be suppressed by using the –exclude_nudge parameter. This default behaviour and the –exclude_nudge parameter have been discontinued. Loose matches of 95% identity or better can now only be listed (i.e., nudged) as Strict matches, regardless of alignment length, by use of the new –include_nudge parameter. As such, these often spurious results are no longer included in default RGI main output.
Also in earlier versions of RGI, by default RGI bwt aligned reads to reference sequences from CARD’s protein homolog models, protein variant models, rRNA mutation models, and protein over-expression models. However, as outlined in the README, the latter three model types require comparison to CARD’s curated lists of mutations known to confer phenotypic antibiotic resistance to differentiate alleles conferring resistance from antibiotic susceptible alleles, e.g. a wild-type gyrase susceptible to fluoroquinolones. As such, earlier versions of RGI were over-reporting antibiotic resistance genes by not checking for these curated mutations. For example, while the KMA algorithm reports SNPs relative to reference, RGI was not screening these SNPs against CARD. Read alignments against the protein variant model, rRNA mutation model, and protein over-expression model reference sequences can now only be listed by use of the new –include_other_models parameter, but at this time these results still do not include comparison to CARD’s curated lists of mutations. As such, these often spurious results are no longer included in default RGI bwt output. Support for mutation screening models will be added to future versions of RGI bwt.