Interpretation logic and output

Interpretation logic and output#

Interpreted AMRFinderplus output#

When given an AMRFinderPlus input file, and an organism, AMRrules will apply the most relevant organism-specific rule to each AMRFinderPlus AMR genotype call. Below is an abbreviated example of the interpreted output file generated by AMRrules. (More examples can be found in Tests).

Abbreviated example of an interpreted AMRFP table

When parsing the AMRFinderPlus input file, AMRrules will first identify what variation type each genotype call belongs to, the gene found, and if there is a mutation in the gene, this mutation will be converted to the AMRrules syntax.

The table below shows the logic of how AMRrules maps AMRFinderPlus genotype calls to the different variation types:

Variation type

AMRFP Method

Additional logic

Gene presence detected

EXACT, ALLELE, BLAST

Protein variant detected

POINTX, POINTP

Nucleotide variant detected

POINTN

Promoter variant detected

POINTN

- symbol present before the nucleotide position

Inactivating mutation detected

INTERNAL_STOP, PARTIAL, POINTX

If POINTX, Subtype must be POINT_DISRUPT

The parsed variation type, gene, and mutation columns are added to the interpreted output file, next to the original AMRFinderPlus columns.

Note

If --print-non-amr is provided to AMRrules, then any additional hits from AMRFinderPlus that are not AMR-related (eg virulence, stress, or metal resistance genes) will also be included in the interpreted output file. These entries will have variation type set to Non-AMR element, and gene and mutation set to -. No rules will be applied to these entries, and all annotation columns will be set to -.

Drug mapping

AMRrules maps the AMRFinderPlus calls to drug or drug class names from the CARD ARO, by parsing the Subclass columns. The dictionary used to map AMRFinderPlus Subclass to CARD ontology is in the AMRrules repository under src/amrrules/resources/amrfp_to_card_drugs_classes.txt.

Some genotype calls have multiple entries in the AMRFinderPlus Subclass field. In such cases, the annotated calls output by AMRrules will contain multiple copies of the genotype row, one for each mapped drug/class. E.g. aac6-Ib-cr is assigned to Subclass “AMIKACIN/KANAMYCIN/QUINOLONE/TOBRAMYCIN” in refgene/AMRfinderplus, so the AMRrules-annotated genotype file will contain 4 rows for this gene, one each for drug = amikacin, drug = tobramycin, drug = kanamycin, and drug class = fluoroquinolone antibiotic

Rule matching#

When matching a genotype call with a rule, the following logic applies:

  1. AMRrules will select all rules matching the variation type.

  2. AMRrules will filter all possible rules to check if the nodeID from the rule matches to Hierarchy node. If not, it will check the parent nodes of Hierarchy node to look for a matching nodeID.

  3. If no matching nodeID is found, AMRrules will then check for matching accessions in the following order:

    1. nucleotide accession

    2. protein accession

    3. HMM accession

For each matching rule, AMRrules appends the following fields from the rules file to the AMRFinderPlus genotype call:

  • ruleID

  • gene context

  • drug

  • drug class

  • phenotype

  • clinical category

  • evidence grade

  • version

  • organism

To get more detailed annotation, set --annot-opts full, which will add the following additional fields from the rule file:

  • breakpoint

  • breakpoint standard

  • breakpoint condition

  • evidence code

  • evidence limitations

  • PMID

  • rule curation note

Note

If multiple rules match to a genotype call, then the genotype call is duplicated, and the relevant information from additional rules is appended. For example, if the genotype call for aac(3) matches to two different rules, one for amikacin and one for kanamycin, then the aac(3) row will be represented twice in the output file, and the rule information for the amikacin will be appended on one row, and kanamycin on the second row.

The resulting output file is stored as <output_prefix>_interpreted.tsv.

Genome summary report#

In addition to the interpreted AMRFinderPlus output file, AMRrules will also generate a genome summary report, stored as <output_prefix>_summary.tsv. This file summarises the resistance per drug/drug class, based on the matched rules.

Some example output files are shown on the installation page.

The following columns are included:

Column

Explanation

sample

Sample ID

drug

Drug, as per CARD ARO. AMRFinderPlus drugs will be converted to CARD ARO format by AMRrules

drug class

Drug class, as per CARD ARO. AMRFinderPlus drugs will be converted to CARD ARO format by AMRrules

clinical category

S/I/R. Highest level of resistance based on markers that match this drug/drug class

phenotype

Wildtype or nonwildtype. Highest level based on markers that match this drug/drug class

evidence grade

None/very low/low/moderate/high. Highest grade of evidence for markers that match this drug/drug class. None is used when we have no evidence for the call, as is the case of rule-less markers.

markers (non-S)

Markers with rules specifying clinical category I or R, separated by ;

markers (no rule)

Markers with no rule, separated by ;

markers (S)

Markers with rules specifying clinical category S, separated by ;

ruleIDs

List of IDs for single-marker rules that apply to this drug/drug class, separated by ;

combo rules

List of IDs multi-marker (combination) rules that apply to this drug/drug class, separated by ;

organism

Organism whose rules were applied

Marker labels

  • Marker labels are formatted as gene:mutation, where the mutation is formatted using AMRrules syntax

  • If the variation type is Gene inactivating mutation the marker will be labelled as gene:- (where the - implies inactivation or functional loss).

  • If the variation type is Gene presence detected, the marker label will be simply gene.

  • If flag-core has been switched on, markers with rules that have gene context == 'core' will be flagged with (core) after the marker. This is the case only if the variation type for that marker is Gene presence detected. eg the core blaSHV gene in K. pneumoniae will now be in the markers (non-S) column as blaSHV-11 (core) for penicillin beta-lactam.

Note

AMRFinderPlus genotype calls detected using POINT_DISRUPT contain detailed information about the disruption, which is formatted using HGVS syntax. For simplicity, by default the AMRrules genome summary report will not show the full mutation after the gene when listing these markers in the genome summary (although they can always be found in the annotated genotype report). Rather, they will be formatted in the genome summary report as gene:- as per other inactivating mutations. If you wish to show the full mutation detected by AMRFinderPlus in the summary report, this can be turned on by providing the option --full-disrupt to the AMRrules call.

Combination Rules#

The genome summary report is where combination rules are applied. For each drug or drug class, the engine extracts all individual rules that have been applied. It then searches the combination rules for this organism and evaluates the logic strings (e.g. “rule1 & rule2 | rule3”) to determine if any apply. If they do, the engine then applies the logic of the combination rule to the call for this drug or drug class.

See details about the rules specification for more information on combination rules and how they are used.

Handling unmatched genotype calls#

By default, if no organism-specific rule is found for a genotype reported by AMRfinderplus, AMRrules will apply an interpretation based on the setting of --no-rule-interpretation. The options for this parameter are:

  • none (default): any unmatched genotype calls will not be assigned to a phenotype or clinical category (entires will be -). Evidence grade will be set to none, as we are not interpreting a phenotype or clinical category.

  • nwt: any unmatched genotype calls will be assigned the phenotype nonwildtype, but no clinical category will be assigned. Evidence grade will be set to none, as we are not interpreting a clinical category.

  • nwtS: any unmatched genotype calls will be assigned the phenotype nonwildtype and clinical category will be set to S. Evidence grade will be set to none, as there is no evidence for the call.

  • nwtR: any unmatched genotype calls will be assigned the phenotype nonwildtype and clinical category will be set to R. Evidence grade will be set to none, as there is no evidence for the call.

To demonstrate the difference between these options, you can compare the output files for the same genome, a MDR Klebsiella pneumoniae strain, using different interpretations for markers with no rules.

Using none (see tests/data/example_output/test_kpneumo_MDR_none_genome_summary.tsv):

Almost all markers have no rules, and so no clinical category, phenotype or evidence grade are set. In the case of the call for penicillin beta-lactam, we have an existing wildtype R rule for blaSHV-1. However, due to the presence of blaTEM-1, which has no rule, we set phenotype to -. However the clinical category of R remains, as regardless of the S/I/R category for the marker with no rule, an R clinical category is still expected.

sample

drug

drug class

clinical category

phenotype

evidence grade

markers (non-S)

markers (no rule)

markers (S)

ruleIDs

combo rules

organism

SAMD00055732

gentamicin

aminoglycoside antibiotic

none

aac(3)-IId;aac(6’)-Ib4

s__Klebsiella pneumoniae

SAMD00055732

streptomycin

aminoglycoside antibiotic

none

aph(3’’)-Ib;aph(6)-Id

s__Klebsiella pneumoniae

SAMD00055732

(all)

carbapenem

none

blaIMP-4

s__Klebsiella pneumoniae

SAMD00055732

trimethoprim

diaminopyrimidine antibiotic

none

dfrA14

s__Klebsiella pneumoniae

SAMD00055732

(all)

fluoroquinolone antibiotic

none

qnrB1

s__Klebsiella pneumoniae

SAMD00055732

ciprofloxacin

fluoroquinolone antibiotic

none

qnrB1

oqxA;oqxB

KPN0002;KPN0003

s__Klebsiella pneumoniae

SAMD00055732

azithromycin

macrolide antibiotic

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

erythromycin

macrolide antibiotic

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

spiramycin

macrolide antibiotic

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

telithromycin

macrolide antibiotic

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

(all)

penicillin beta-lactam

R

none

blaSHV-1

blaTEM-1

KPN0001

s__Klebsiella pneumoniae

SAMD00055732

fosfomycin

phosphonic acid antibiotic

S

wildtype

moderate

fosA

KPN0004

s__Klebsiella pneumoniae

SAMD00055732

(all)

sulfonamide antibiotic

none

sul2;sul1

s__Klebsiella pneumoniae

SAMD00055732

(all)

third-generation cephalosporin

none

blaCTX-M-15;blaOXA-1

s__Klebsiella pneumoniae

SAMD00055732

(n/a)

antibiotic efflux

none

emrD

s__Klebsiella pneumoniae

SAMD00055732

catB3:-;aac(3)-IIe:-

none (partial hits)

s__Klebsiella pneumoniae

Using nwt (see tests/data/example_output/test_kpneumo_MDR_nwt_genome_summary.tsv):

Now, all drugs with markers that have no rules have nonwildtype as their phenotype. This includes the call for penicillin beta-lactam. Evidence grades are set to none as we have no evidence for these calls.

sample

drug

drug class

clinical category

phenotype

evidence grade

markers (non-S)

markers (no rule)

markers (S)

ruleIDs

combo rules

organism

SAMD00055732

gentamicin

aminoglycoside antibiotic

nonwildtype

none

aac(3)-IId;aac(6’)-Ib4

s__Klebsiella pneumoniae

SAMD00055732

streptomycin

aminoglycoside antibiotic

nonwildtype

none

aph(3’’)-Ib;aph(6)-Id

s__Klebsiella pneumoniae

SAMD00055732

(all)

carbapenem

nonwildtype

none

blaIMP-4

s__Klebsiella pneumoniae

SAMD00055732

trimethoprim

diaminopyrimidine antibiotic

nonwildtype

none

dfrA14

s__Klebsiella pneumoniae

SAMD00055732

(all)

fluoroquinolone antibiotic

nonwildtype

none

qnrB1

s__Klebsiella pneumoniae

SAMD00055732

ciprofloxacin

fluoroquinolone antibiotic

nonwildtype

none

qnrB1

oqxA;oqxB

KPN0002;KPN0003

s__Klebsiella pneumoniae

SAMD00055732

azithromycin

macrolide antibiotic

nonwildtype

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

erythromycin

macrolide antibiotic

nonwildtype

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

spiramycin

macrolide antibiotic

nonwildtype

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

telithromycin

macrolide antibiotic

nonwildtype

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

(all)

penicillin beta-lactam

R

nonwildtype

none

blaSHV-1

blaTEM-1

KPN0001

s__Klebsiella pneumoniae

SAMD00055732

fosfomycin

phosphonic acid antibiotic

S

wildtype

moderate

fosA

KPN0004

s__Klebsiella pneumoniae

SAMD00055732

(all)

sulfonamide antibiotic

nonwildtype

none

sul2;sul1

s__Klebsiella pneumoniae

SAMD00055732

(all)

third-generation cephalosporin

nonwildtype

none

blaCTX-M-15;blaOXA-1

s__Klebsiella pneumoniae

SAMD00055732

(n/a)

antibiotic efflux

nonwildtype

none

emrD

s__Klebsiella pneumoniae

SAMD00055732

catB3:-;aac(3)-IIe:-

none (partial hits)

s__Klebsiella pneumoniae

Using nwtS (see tests/data/example_output/test_kpneumo_MDR_nwtS_genome_summary.tsv):

Now, all drugs with markers that have no rules have S as their clinical category, and nonwildtype as their phenotype. As an S call for blaTEM-1 doesn’t supercede the existing R rule for the blaSHV-1 marker, this clinical category remains the same. Again, evidence grades are set to none as we have no evidence for these calls.

sample

drug

drug class

clinical category

phenotype

evidence grade

markers (non-S)

markers (no rule)

markers (S)

ruleIDs

combo rules

organism

SAMD00055732

gentamicin

aminoglycoside antibiotic

S

nonwildtype

none

aac(3)-IId;aac(6’)-Ib4

s__Klebsiella pneumoniae

SAMD00055732

streptomycin

aminoglycoside antibiotic

S

nonwildtype

none

aph(3’’)-Ib;aph(6)-Id

s__Klebsiella pneumoniae

SAMD00055732

(all)

carbapenem

S

nonwildtype

none

blaIMP-4

s__Klebsiella pneumoniae

SAMD00055732

trimethoprim

diaminopyrimidine antibiotic

S

nonwildtype

none

dfrA14

s__Klebsiella pneumoniae

SAMD00055732

(all)

fluoroquinolone antibiotic

S

nonwildtype

none

qnrB1

s__Klebsiella pneumoniae

SAMD00055732

ciprofloxacin

fluoroquinolone antibiotic

S

nonwildtype

none

qnrB1

oqxA;oqxB

KPN0002;KPN0003

s__Klebsiella pneumoniae

SAMD00055732

azithromycin

macrolide antibiotic

S

nonwildtype

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

erythromycin

macrolide antibiotic

S

nonwildtype

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

spiramycin

macrolide antibiotic

S

nonwildtype

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

telithromycin

macrolide antibiotic

S

nonwildtype

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

(all)

penicillin beta-lactam

R

nonwildtype

none

blaSHV-1

blaTEM-1

KPN0001

s__Klebsiella pneumoniae

SAMD00055732

fosfomycin

phosphonic acid antibiotic

S

wildtype

moderate

fosA

KPN0004

s__Klebsiella pneumoniae

SAMD00055732

(all)

sulfonamide antibiotic

S

nonwildtype

none

sul2;sul1

s__Klebsiella pneumoniae

SAMD00055732

(all)

third-generation cephalosporin

S

nonwildtype

none

blaCTX-M-15;blaOXA-1

s__Klebsiella pneumoniae

SAMD00055732

(n/a)

antibiotic efflux

none

emrD

s__Klebsiella pneumoniae

SAMD00055732

catB3:-;aac(3)-IIe:-

none (partial hits)

s__Klebsiella pneumoniae

Using nwtR (see tests/data/example_output/test_kpneumo_MDR_nwtR_genome_summary.tsv):

Finally, all drugs with markers that have no rules have R as their clinical category, and nonwildtype as their phenotype. Evidence grades are set to none as we have no evidence for these calls.

sample

drug

drug class

clinical category

phenotype

evidence grade

markers (non-S)

markers (no rule)

markers (S)

ruleIDs

combo rules

organism

SAMD00055732

gentamicin

aminoglycoside antibiotic

R

nonwildtype

none

aac(3)-IId;aac(6’)-Ib4

s__Klebsiella pneumoniae

SAMD00055732

streptomycin

aminoglycoside antibiotic

R

nonwildtype

none

aph(3’’)-Ib;aph(6)-Id

s__Klebsiella pneumoniae

SAMD00055732

(all)

carbapenem

R

nonwildtype

none

blaIMP-4

s__Klebsiella pneumoniae

SAMD00055732

trimethoprim

diaminopyrimidine antibiotic

R

nonwildtype

none

dfrA14

s__Klebsiella pneumoniae

SAMD00055732

(all)

fluoroquinolone antibiotic

R

nonwildtype

none

qnrB1

s__Klebsiella pneumoniae

SAMD00055732

ciprofloxacin

fluoroquinolone antibiotic

R

nonwildtype

none

qnrB1

oqxA;oqxB

KPN0002;KPN0003

s__Klebsiella pneumoniae

SAMD00055732

azithromycin

macrolide antibiotic

R

nonwildtype

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

erythromycin

macrolide antibiotic

R

nonwildtype

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

spiramycin

macrolide antibiotic

R

nonwildtype

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

telithromycin

macrolide antibiotic

R

nonwildtype

none

mph(A)

s__Klebsiella pneumoniae

SAMD00055732

(all)

penicillin beta-lactam

R

nonwildtype

high

blaSHV-1

blaTEM-1

KPN0001

s__Klebsiella pneumoniae

SAMD00055732

fosfomycin

phosphonic acid antibiotic

S

wildtype

moderate

fosA

KPN0004

s__Klebsiella pneumoniae

SAMD00055732

(all)

sulfonamide antibiotic

R

nonwildtype

none

sul2;sul1

s__Klebsiella pneumoniae

SAMD00055732

(all)

third-generation cephalosporin

R

nonwildtype

none

blaCTX-M-15;blaOXA-1

s__Klebsiella pneumoniae

SAMD00055732

(n/a)

antibiotic efflux

emrD

s__Klebsiella pneumoniae

SAMD00055732

catB3:-;aac(3)-IIe:-

none (partial hits)

s__Klebsiella pneumoniae
Contents