A deep learning based approach to predict Antibiotic Resistance Genes (ARGs) from metagenomes. It provides two models,deepARG-SS and deepARG-LS.
- updated on Nov 10 - 2023
- deeparg 1.0.2: Added to pip
- Fastq input - short reads pipeline fixed
DeepARG generates two files: *.ARG that contains the sequences with a probability >= --prob (0.8 default) and *.potential.ARG with sequences containing a probability < --prob (0.8 default). The *.potential.ARG file can still contain ARG-like sequences, howevere, it is necessary inspect its sequences.
The output format for both files consists of the following fields:
* ARG_NAME
* QUERY_START
* QUERY_END
* QUERY_ID
* PREDICTED_ARG_CLASS
* BEST_HIT_FROM_DATABASE
* PREDICTION_PROBABILITY
* ALIGNMENT_BESTHIT_IDENTITY (%)
* ALIGNMENT_BESTHIT_LENGTH
* ALIGNMENT_BESTHIT_BITSCORE
* ALIGNMENT_BESTHIT_EVALUE
* COUNTS
DeepARG is under Python 2.7, therefore, it is recommended to run it via virtual environment or via docker.
Install miniconda https://docs.conda.io/en/latest/miniconda.html
Create a virtual environment with conda:
conda create -n deeparg_env python=2.7.18
source activate deeparg_env
Install diamond with conda (inside virtual environment):
conda install -c bioconda diamond==0.9.24
Optional (used for short reads pipeline):
conda install -c bioconda trimmomatic
conda install -c bioconda vsearch
conda install -c bioconda bedtools==2.29.2
conda install -c bioconda bowtie2==2.3.5.1
conda install -c bioconda samtools
Install deeparg with pip and download the data required by deeparg
pip install git+https://github.com/gaarangoa/deeparg.git
deeparg download_data -o /path/to/local/directory/
Activate virtual environment
conda activate deeparg_env
Deactivate the virtual environment:
conda deactivate
In this example, we will classify a set of ORFs from a set of assembled contigs. The fasta file contains gene sequences (nucleotides).
deeparg predict \
--model LS \
-i ./test/ORFs.fa \
-o ./test/X \
-d /path/to/data/ \
--type nucl \
--min-prob 0.8 \
--arg-alignment-identity 30 \
--arg-alignment-evalue 1e-10 \
--arg-num-alignments-per-entry 1000
usage: deeparg predict
-h, --help show this help message and exit
--model MODEL Select model to use (short sequences for reads | long
sequences for genes) SS|LS [No default]
-i INPUT_FILE, --input-file INPUT_FILE
Input file (Fasta input file)
-o OUTPUT_FILE, --output-file OUTPUT_FILE
Output file where to store results
-d DATA_PATH, --data-path DATA_PATH
Path where data was downloaded [see deeparg download-
data --help for details]
--type TYPE Molecular data type prot/nucl [Default: nucl]
--min-prob MIN_PROB Minimum probability cutoff [Default: 0.8]
--arg-alignment-identity ARG_ALIGNMENT_IDENTITY
Identity cutoff for sequence alignment [Default: 50]
--arg-alignment-evalue ARG_ALIGNMENT_EVALUE
Evalue cutoff [Default: 1e-10]
--arg-alignment-overlap ARG_ALIGNMENT_OVERLAP
Alignment read overlap [Default: 0.8]
--arg-num-alignments-per-entry ARG_NUM_ALIGNMENTS_PER_ENTRY
Diamond, minimum number of alignments per entry
[Default: 1000]
--model-version MODEL_VERSION
Model deepARG version [Default: v2]
Go to the deeparg-ss directory and run any of the following commands:
Input is a FASTA file:
1) Annotate gene-like sequences when the input is a nucleotide FASTA file:
deeparg predict --model LS --type nucl --input /path/file.fasta --out /path/to/out/file.out
2) Annotate gene-like sequences when the input is an amino acid FASTA file:
deeparg predict --model LS --type prot --input /path/file.fasta --out /path/to/out/file.out
3) Annotate short sequence reads when the input is a nucleotide FASTA file:
deeparg predict --model SS --type nucl --input /path/file.fasta --out /path/to/out/file.out
3) Annotate short sequence reads when the input is a protein FASTA file (unusual case):
deeparg predict --model SS --type prot --input /path/file.fasta --out /path/to/out/file.out
deeparg short_reads_pipeline [-h] --forward_pe_file FORWARD_PE_FILE
--reverse_pe_file REVERSE_PE_FILE
--output_file OUTPUT_FILE
[-d DEEPARG_DATA_PATH]
[--deeparg_identity DEEPARG_IDENTITY]
[--deeparg_probability DEEPARG_PROBABILITY]
[--deeparg_evalue DEEPARG_EVALUE]
[--gene_coverage GENE_COVERAGE]
[--bowtie_16s_identity BOWTIE_16S_IDENTITY]
optional arguments:
-h, --help show this help message and exit
--forward_pe_file FORWARD_PE_FILE
forward mate from paired end library
--reverse_pe_file REVERSE_PE_FILE
reverse mate from paired end library
--output_file OUTPUT_FILE
save results to this file prefix
-d DEEPARG_DATA_PATH, --deeparg_data_path DEEPARG_DATA_PATH
Path where data was downloaded [see deeparg download-
data --help for details]
--deeparg_identity DEEPARG_IDENTITY
minimum identity for ARG alignments [default 80]
--deeparg_probability DEEPARG_PROBABILITY
minimum probability for considering a reads as ARG-
like [default 0.8]
--deeparg_evalue DEEPARG_EVALUE
minimum e-value for ARG alignments [default 1e-10]
--gene_coverage GENE_COVERAGE
minimum coverage required for considering a full gene
in percentage. This parameter looks at the full gene
and all hits that align to the gene. If the overlap of
all hits is below the threshold the gene is discarded.
Use with caution [default 1]
deeparg short_reads_pipeline \
--forward_pe_file ./test/F.fq.gz \
--reverse_pe_file ./test/R.fq.gz \
--output_file ./test/reads \
-d ~/Desktop/darg \
--bowtie_16s_identity 100
If you use deepARG in published research, please cite:
Arango-Argoty GA, Garner E, Pruden A, Heath LS, Vikesland P, Zhang L. DeepARG: A deep learning approach for predicting antibiotic resistance genes from metagenomic data. Microbiome20186:23 https://doi.org/10.1186/s40168-018-0401-z.
Database is hosted in Zenodo: https://zenodo.org/records/8280582
deepARG is under the MIT licence. However, please take a look at te comercial restrictions of the databases used during the mining process (CARD, ARDB, and UniProt).
If need any asistance please contact: [email protected]