Ranking the risk of antibiotic resistance for genomes/metagenomes

antibiotic, resistance, risk, one, health, clinical, AMR, mobile, antibiotic-resistance, metagenomics, risk-assessment
pip install arg-ranker==2.8.6



arg_ranker evaluates the risk of ARGs in genomes and metagenomes


pip install arg_ranker


  • python 3
  • diamond: conda install -c bioconda diamond (
  • blast+: conda install -c bioconda blast (
  • For metagenomes:
    • kraken2: conda install -c bioconda kraken2(
      • to compute the abundance of ARGs as copy number of ARGs per bacterial cell (recommended)
        • download the kraken2 standard database (50 GB of disk space): kraken2-build --standard --db $KRAKENDB
          where $KRAKENDB is your preferred database name/location
        • MicrobeCensu: git clone && cd MicrobeCensus && python install to estimate the average genome size for metagenomes. (
      • to compute the abundance of ARGs as copy number of ARGs per 16S
        • download the kraken2 16S database (73.2 MB of disk space): kraken2-build --db $DBNAME --special greengenes

How to use it

  • put all your genomes (.fa or .fasta) and metagenomes (.fq or .fastq) into one folder ($INPUT)
  • run arg_ranker -i $INPUT (genomes only)
  • run arg_ranker -i $INPUT -kkdb $KRAKENDB (genomes/metagenomes + kraken2 standard database)
    • or run arg_ranker -i $INPUT -kkdb $KRAKENDB -kkdbtype 16S (kraken2 16S database)
  • run sh arg_ranking/script_output/


  • Sample_ranking_results.txt (Table 1)

    Sample Rank_I_per Rank_II_per Rank_III_per Rank_IV_per Unassessed_per Total_abu Rank_code Rank_I_risk Rank_II_risk Rank_III_risk Rank_IV_risk ARGs_unassessed_risk note1
    WEE300_all-trimmed-decont_1.fastq 4.6E-02 0.0E+00 6.8E-02 7.5E-01 1.3E-01 5.4E-04 1.0-0.0-0.5-1.7-0.3 1.0 0.0 0.5 1.7 0.3 hospital_metagenome
    EsCo_genome.fasta 0.0E+00 0.0E+00 0.0E+00 1.0E+00 0.0E+00 2.0E+00 0.0-0.0-0.0-2.2-0.0 0.0 0.0 0.0 2.2 0.0 E.coli_genome
  1. Rank_I_per - Unassessed_per: percentage of ARGs of a risk Rank
    Total_abu: total abundance of all ARGs
  2. For genomes, we output the copy number of ARGs detected in each genome.
  3. For metagenomes, we compute the abundance of ARGs as the copy number of ARGs divided by the bacterial cell number or 16S copy number in the same metagenome.
    If you downloaded the kraken2 standard database, we compute the copy number of ARGs divided by the bacterial cell number.
    If you downloaded the kraken2 16S database, we compute the copy number of ARGs divided by the 16S copy number.
    The copy number of ARGs, 16S, and bacterial cells were computed as the number of reads mapped to them divided by their gene/genome length.
  4. We compute the contribution of each ARG risk Rank as the average abundance of ARGs of a risk Rank divided by the average abundance of all ARGs
    Rank_I_risk - Unassessed_risk: the contribution of ARGs of a risk Rank
    Rank_code: a code of contribution from Rank I to Unassessed
  • Sample_ARGpresence.txt:
    The abundance, the gene family, and the antibiotic of resistance of ARGs detected in the input samples


run arg_ranker -i example -kkdb $KRAKENDB
run sh arg_ranking/script_output/
The arg_ranking/Sample_ranking_results.txt should look like Table 1 (using kraken2 standard database)

Metadata for your samples (optional)

arg_ranker can merge your sample metadata into the results of ARG ranking (i.e. note1 in Table 1).
Simply put all information you would like to include into a tab-delimited table
Make sure that your sample names are listed as the first column (check example/metadata.txt).


Dr. An-Ni Zhang (MIT), Prof. Eric Alm (MIT), Prof. Tong Zhang* (University of Hong Kong)


Zhang, AN., Gaston, J.M., Dai, C.L. et al. An omics-based framework for assessing the health risk of antimicrobial resistance genes. Nat Commun 12, 4765 (2021).

Contact or