geNomad: Identification of mobile genetic elements
geNomad's primary goal is to identify viruses and plasmids in sequencing data (isolates, metagenomes, and metatranscriptomes). It also provides a couple of additional features that can help you in your analysis:
- Taxonomic assignment of viral genomes.
- Identification of viruses integrated in host genomes (proviruses).
- Functional annotation of proteins.
For installation instructions, information about how geNomad works, and a detailed explanation of how to execute it, please check the full documentation: https://portal.nersc.gov/genomad/
We recommend users to read the documentation before starting to use geNomad. If you are in a rush, however, you can follow this quick step-by-step example.
First, you need to install geNomad. There's a couple of ways to do that, but here we will use conda as it will handle all dependencies for us.
# Create a conda environment for geNomad
conda create -n genomad -c conda-forge -c bioconda genomad
# Activate the geNomad environment
conda activate genomad
geNomad depends on a database that contains the profiles of the markers that are used to classify sequences, their taxonomic information, their functional annotation, etc. So, you should first download the database to your current directory:
genomad download-database .
The database will be contained within the genomad_db directory.
Now you are ready to go! geNomad works by executing a series of modules sequentially (more on that in the documentation), but we provide a convenient end-to-end command that will execute the entire pipeline for you in one go.
In this example, we will use an Escherichia coli genome (GCF_000008865.2) as input. You can use any FASTA file containing nucleotide sequences as input. geNomad will work for isolate genomes, metagenomes, and metatranscriptomes.
The command to execute geNomad is structured like this:
genomad end-to-end [OPTIONS] INPUT OUTPUT DATABASE
So, to run the full geNomad pipeline (end-to-end command), taking a nucleotide FASTA file (GCF_000008865.2.fna.gz) and the database (genomad_db) as input, we will execute the following command:
genomad end-to-end --min-score 0.8 --cleanup --splits 16 GCF_000008865.2.fna.gz genomad_output genomad_db
The results will be written inside the genomad_output directory.
Three important details about the command above:
- By setting
--min-scoreto0.8we make the classification more conservative, as only sequences with a virus/plasmid score higher than 0.8 will be reported. - The
--cleanupoption was used to force geNomad to delete intermediate files that were generated during the execution. This will save you some storage space. - The
--splits 16parameter was used here to make it possible to run this example in a notebook. geNomad searches a big database of protein profiles that take up a lot of space in memory. To prevent the execution from failing due to insufficient memory, we can use the--splitsparameter to split the seach into chuncks. If you are running geNomad in a big server you might not need to split your search, increasing the execution speed. - Note that the input FASTA file that I used as input was compressed. This is possible because geNomad supports input files compressed as
.gz,.bz2, or.xz.
In this example, the results of geNomad's analysis will be written to the genomad_output directory, which will look like this:
genomad_output
├── GCF_000008865.2_aggregated_classification
├── GCF_000008865.2_aggregated_classification.log
├── GCF_000008865.2_annotate
├── GCF_000008865.2_annotate.log
├── GCF_000008865.2_find_proviruses
├── GCF_000008865.2_find_proviruses.log
├── GCF_000008865.2_marker_classification
├── GCF_000008865.2_marker_classification.log
├── GCF_000008865.2_nn_classification
├── GCF_000008865.2_nn_classification.log
├── GCF_000008865.2_summary
╰── GCF_000008865.2_summary.log
As mentioned above, geNomad works by executing several modules sequentially. Each one of these will produce a log file (<prefix>_<module>.log) and a subdirectory (<prefix>_<module>).
For this example, we will only look at the files within GCF_000008865.2_summary. The <prefix>_summary directory contains files that summarize the results that were generated across the pipeline. If you just want a list of the plasmids and viruses identified in your input, this is what you are looking for.
genomad_output
╰── GCF_000008865.2_summary
├── GCF_000008865.2_plasmid.fna
├── GCF_000008865.2_plasmid_genes.tsv
├── GCF_000008865.2_plasmid_proteins.faa
├── GCF_000008865.2_plasmid_summary.tsv
├── GCF_000008865.2_summary.json
├── GCF_000008865.2_virus.fna
├── GCF_000008865.2_virus_genes.tsv
├── GCF_000008865.2_virus_proteins.faa
╰── GCF_000008865.2_virus_summary.tsv
First, let's look at GCF_000008865.2_virus_summary.tsv:
seq_name length topology coordinates n_genes genetic_code virus_score fdr n_hallmarks marker_enrichment taxonomy
------------------------------------ ------ -------- --------------- ------- ------------ ----------- --- ----------- ----------------- ---------------------------------------------------------------
NC_002695.2|provirus_1245607_1309390 63784 Provirus 1245607-1309390 90 11 0.9735 NA 11 109.6197 Viruses;Duplodnaviria;Heunggongvirae;Uroviricota;Caudoviricetes
NC_002695.2|provirus_891197_928364 37168 Provirus 891197-928364 42 11 0.9704 NA 19 49.6673 Viruses;Duplodnaviria;Heunggongvirae;Uroviricota;Caudoviricetes
NC_002695.2|provirus_5041220_5079596 38377 Provirus 5041220-5079596 53 11 0.9698 NA 22 60.6887 Viruses;Duplodnaviria;Heunggongvirae;Uroviricota;Caudoviricetes
…
This tabular file lists all the viruses that geNomad found in your input and gives you some convenient information about them. Here's what each column contains:
seq_name: The identifier of the sequence in the input FASTA file. Proviruses will have the following name scheme:<sequence_identifier>|provirus_<start_coordinate>_<end_coordinate>.length: Length of the sequence (or the provirus, in the case of integrated viruses).topology: Topology of the viral sequence. Possible values are: Linear, DTR (direct terminal repeats), ITR (inverted terminal repeats), or Provirus (viruses integrated in host genomes).coordinates: 1-indexed coordinates of the provirus region within host sequences. Will beNAfor viruses that were not predicted to be integrated.n_genes: Number of genes encoded in the sequence.genetic_code: Predicted genetic code. Possible values are: 11 (standard code for Bacteria and Archaea), 4 (recoded TGA stop codon), or 15 (recoded TAG stop codon).virus_score: A measure of how confident geNomad is that the sequence is a virus. Sequences that have scores close to 1.0 are more likely to be viruses than the ones that have lower scores.fdr: The estimated false discovery rate (FDR) of the classification (that is, the expected proportion of false positives among the sequences up to this row). To estimate FDRs geNomad requires score calibration, which is turned off by default. Therefore, this column will only containNAvalues in this example.n_hallmarks: Number of genes that matched a hallmark geNomad marker. Hallmarks are genes that were previously associated with viral function and their presence is a strong indicative that the sequence is indeed a virus.marker_enrichment: A score that represents the total enrichment of viral markers in the sequence. The value goes as the number of virus markers in the sequence increases, so sequences with multiple markers will have higher score. Chromosome and plasmid markers will reduce the score.taxonomy: Taxonomic assignment of the virus genome. Lineages follow the taxonomy contained in ICTV's VMR number 19.
In our example, geNomad identified several proviruses integrated into the E. coli genome. They were all predicted to use the genetic code 11 and were assigned to the Caudoviricetes class, which contains all the tailed bacteriphages. Since they all have high scores, we can be confident that these are indeed viruses.
Another important file is GCF_000008865.2_virus_genes.tsv. During its execution, geNomad annotates the genes encoded by the input sequences using a database of chromosome, plasmid, and virus-specific markers. The <prefix>_virus_genes.tsv file summarizes the annotation of the genes encoded by the identified viruses.
gene start end length strand gc_content genetic_code rbs_motif marker evalue bitscore uscg plasmid_hallmark virus_hallmark taxid taxname annotation_conjscan annotation_accessions annotation_description
-------------------------------------- ------ ------ ------ ------ ---------- ------------ -------------- ----------------- --------- -------- ---- ---------------- -------------- ----- -------------- ------------------- ----------------------------------- ----------------------------------------------
NC_002695.2|provirus_300073_325822_264 300073 301047 975 -1 0.476 11 AGGAG/GGAGG NA NA NA 0 0 0 1 NA NA NA NA
NC_002695.2|provirus_300073_325822_265 301423 301812 390 -1 0.392 11 GGA/GAG/AGG NA NA NA 0 0 0 1 NA NA NA NA
NC_002695.2|provirus_300073_325822_266 301940 302653 714 -1 0.461 11 None GENOMAD.222303.VP 4.577e-09 59 0 0 0 2561 Caudoviricetes NA PF06223.15;K10762 Minor tail protein T
NC_002695.2|provirus_300073_325822_267 302754 302954 201 1 0.428 11 AGGAGG GENOMAD.061471.VV 1.653e-15 71 0 0 1 2561 Caudoviricetes NA PF09048.13;TIGR03339;K22302;COG1609 Cro
NC_002695.2|provirus_300073_325822_268 303073 303366 294 1 0.503 11 AGGA/GGAG/GAGG GENOMAD.129061.VV 5.21e-33 123 0 0 1 2561 Caudoviricetes NA PF05269.14;TIGR00721 Bacteriophage CII protein
…
The columns in this file are:
gene: Identifier of the gene (<sequence_name>_<gene_number>). Gene numbers start with 1 (first gene in the sequence). Because the genes in this example are encoded by prophages integrated in the middle of the host chromosome, they won't necessarily start with 1.start: 1-indexed start coordinate of the gene.end: 1-indexed end coordinate of the gene.length: Length of the gene locus (in base pairs).strand: Strand that encodes the gene. Can be 1 (direct strand) or -1 (reverse strand).gc_content: GC content of the gene locus.genetic_code: Predicted genetic code (see details in the explanation of the summary file).rbs_motif: Detected motif of the ribosome-binding site.marker: Best matching geNomad marker. If this gene doesn't match any markers, the value will beNA.evalue: E-value of the alignment between the protein encoded by the gene and the best matching geNomad marker.bitscore: Bitscore of the alignment between the protein encoded by the gene and the best matching geNomad marker.uscg: Whether the marker assigned to this gene corresponds to a universal single copy gene (UCSG). These genes are expected to be found in chromosomes and are rare in plasmids and viruses. Can be 1 (gene is USCG) or 0 (gene is not USCG).plasmid_hallmark: Whether the marker assigned to this gene represents a plasmid hallmark.virus_hallmark: Whether the marker assigned to this gene represents a virus hallmark.taxid: Taxonomic identifier of the marker assigned to this gene (you can ignore this as it is meant to be used internally by geNomad).taxname: Name of the taxon associated with the assigned geNomad marker. In this example, we can see that the annotated proteins are all characteristic of Caudoviricetes (which is why the provirus was assigned to this class).annotation_conjscan: If the marker that matched the gene is a conjugation-related gene (as defined in CONJscan) this field will show which CONJscan acession was assigned to the marker.annotation_accessions: Some of the geNomad markers are functionally annotated. This column tells you which entries in Pfam, TIGRFAM, COG, and KEGG were assigned to the marker.annotation_description: A text describing the function assigned to the marker.
In the snippet above we can see the information of the first five genes encoded by NC_002695.2|provirus_300073_325822. The first two entried didn't match any geNomad marker. The following three were all assigned to protein families that are typical of tailed bacteriphages (such as the minor tail protein), which reassures us that these are indeed Caudoviricetes.
One important detail here is that the primary purpose of geNomad's markers is classification. They were designed to be specific to chromosomes, plasmids, or viruses, enabling the distinction of sequences belonging to these classes. Therefore, you should not expect that every single viral gene will be annotated with a geNomad marker. If you want to annotate the genes within your sequences as throughly as possible, you should use databases such as Pfam or EggNOG.
The other two virus-related files within the summary directory are GCF_000008865.2_virus.fna and CF_000008865.2_virus_proteins.faa. These are FASTA files of the identified virus sequences and their proteins, respectively. Proviruses are automatically excised from the host sequence.
Enough with viruses. What about the plasmids?
As you would expect, the data pertaining to the identification of plasmids can be found in the <prefix>_plasmid_summary.tsv, <prefix>_genes.tsv, <prefix>_plasmid.fna, and <prefix>_plasmid_proteins.faa files. These are mostly very similar to their virus counterparts. The differences in <prefix>_plasmid_summary.tsv (shown below) are the following:
- Virus-specific columns that are in
<prefix>_virus_summary.tsv(coordinatesandtopology) are not present. - The
conjugation_genescolumn lists genes that might involved in conjugation. It's important to note that the presence of such genes is not sufficient to tell whether a given plasmid is conjugative or mobilizible. If you are interested in identifying conjugative plasmids, we recommend you to analyze the plasmids you identified using geNomad with CONJscan.
seq_name length topology n_genes genetic_code plasmid_score fdr n_hallmarks marker_enrichment conjugation_genes
----------- ------ -------- ------- ------------ ------------- --- ----------- ----------------- -----------------
NC_002128.1 92721 Linear 88 11 0.9942 NA 5 46.4458 T_virB11;MOBP1
NC_002127.1 3306 Linear 3 11 0.9913 NA 1 1.6586 NA