Cenote-Taker3

Discover and annotate the virome.

Works on your laptop or HPC (compatible with MacOS and Linux)

Cenote-Taker 3 is a virus bioinformatics tool that scales from individual genomes sequences to massive metagenome assemblies to:

1. Identify sequences containing genes specific to viruses (virus hallmark genes)

2. Annotate virus sequences including:

---a) adaptive ORF calling

---b) a large catalog of HMMs from virus gene families for functional annotation

---c) Hierarchical taxonomy assignment based on hallmark genes

---d) mmseqs2-based CDD database search

---e) tabular (.tsv) and interactive genome map (.gbf) outputs

Also, Cenote-Taker 3 is very fast, many many times faster than Cenote-Taker 2 for large datasets, and faster than comparable annotation using pharokka with more function annotation for virus genes (in my hands)

Image of example genome map:

Use Cases

1. Discovering virus contigs in metagenomic data

2. Annotating virus sequences without highly similar well-annotated reference

3. Finding prophages (or proviruses) in microbial genomes

Not-Use Cases

1. Not for read-level classification of known viruses (see Marker-MAGu or EsViritu for this task)

2. Not ideal for annotating virus genomes that are highly similar to known references (e.g. phage lambda with a few mutations).

Schematic

Installation Instructions

Most recent versions

Cenote-Taker 3 scripts: v3.4.3 Cenote-Taker 3 Databases: v3.1.1

This should work on MacOS and Linux

Versions used in test installations

mamba 1.5.8

conda 24.7.1

Bioconda package (most users)

mamba is better/faster than conda for almost all solving/installation tasks

1. Use mamba to install the bioconda package

macOS (specify osx-64 platform regardless of which chip you have) I'm also noticing a macOS-specific issue with newer mmseqs versions, so use mmseqs2=15.6f452

mamba create --platform osx-64 -n ct3_env -c conda-forge -c bioconda cenote-taker3=3.4.3 mmseqs2=15.6f452

linux

mamba create -n ct3_env -c conda-forge -c bioconda cenote-taker3=3.4.3

Using conda instead

macOS (specify osx-64 platform regardless of which chip you have)

conda create --platform osx-64 -n ct3_env -c conda-forge -c bioconda cenote-taker3=3.4.3 mmseqs2=15.6f452

linux

conda create -n ct3_env -c conda-forge -c bioconda cenote-taker3=3.4.3

2. Activate the conda environment.

conda activate ct3_env

You should be able to type cenotetaker3 and get_ct3_dbs in terminal to bring up help menu now

3. Change to a directory where you'd like to install databases and run database script, specify DB directory with -o.

Total DB file size of 3.0 GB after file decompression

cd ..

get_ct3_dbs -o ct3_DBs --hmm T --hallmark_tax T --refseq_tax T --mmseqs_cdd T --domain_list T

With optional hhsuite databases

Warning: due to inconsistent server speed, these downloads may take over 2 hours.

You may download one or more hhsuite DB.

The data footprint is:

Database	Size
CDD	6.1 GB
pfam	4.6 GB
pdb70	56 GB

get_ct3_dbs -o ct3_DBs --hmm T --hallmark_tax T --refseq_tax T --mmseqs_cdd T --domain_list T --hhCDD T --hhPFAM T --hhPDB T

4. Set the database directory as a conda environmental variable.

conda env config vars set CENOTE_DBS=/path/to/ct3_DBs

From source (development versions)

1. Clone this GitHub repo

2. Using mamba (package manager within conda) and the provided yaml file, make the environment:

mamba env create -f Cenote-Taker3/environment/ct3_env.yaml

3. Activate the conda environment.

conda activate ct3_env

4. Change to repo and pip install command line tool.

cd Cenote-Taker3

pip install .

You should be able to type cenotetaker3 and get_ct3_dbs in terminal to bring up help menu now

5. Change to a directory where you'd like to install databases and run database script, specify DB directory with -o.

Total DB file size of 3.0 GB after file decompression

cd ..

get_ct3_dbs -o ct3_DBs --hmm T --hallmark_tax T --refseq_tax T --mmseqs_cdd T --domain_list T

With optional hhsuite databases

Warning: due to inconsistent server speed, these downloads may take over 2 hours.

You may download one or more hhsuite DB.

The data footprint is:

Database	Size
CDD	6.1 GB
pfam	4.6 GB
pdb70	56 GB

get_ct3_dbs -o ct3_DBs --hmm T --hallmark_tax T --refseq_tax T --mmseqs_cdd T --domain_list T --hhCDD T --hhPFAM T --hhPDB T

6. Set the database directory as a conda environmental variable.

conda env config vars set CENOTE_DBS=/path/to/ct3_DBs

Running Cenote-Taker 3

Make sure conda environment is activated

Help Menu

cenotetaker3 -h

Test contigs

cenotetaker3 -c Cenote-Taker3/test_data/testcontigs_DNA_ct2.fasta -r test_ct3 -p T

Default Discover and Annotate

cenotetaker3 -c my_metagenome_contigs.fna -r my_meta_ct3 -p T

Recommended settings for microbial genomes

cenotetaker3 -c my_metagenome_contigs.fna -r my_meta_ct3 -p T --lin_minimum_hallmark_genes 2

Discover and Annotate, Force prodigal (prodigal-gv is default)

cenotetaker3 -c my_metagenome_contigs.fna -r my_meta_ct3pr -p T --caller prodigal

Just Annotate

cenotetaker3 -c my_virus_contigs.fna -r my_virs_ct3 -p F -am T

Choose which HMM DBs are hallmark (virion rdrp is default)

cenotetaker3 -c my_metagenome_contigs.fna -r my_meta_ct3 -p T -db virion rdrp dnarep

Calculate coverage level with reads

cenotetaker3 -c my_metagenome_contigs.fna -r my_meta_ct3 -p T --reads my_reads/*fastq

Output Files

{run_title}/
|   {run_title}_virus_summary.tsv                 <- main summary file for each virus
|   {run_title}_virus_sequences.fna               <- all virus genome seqs
|   {run_title}_virus_AA.faa                      <- all virus AA seqs
|   {run_title}_prune_summary.tsv                 <- summary of pruning of each sequence
|   final_genes_to_contigs_annotation_summary.tsv <- annotation info, all genes
|   run_arguments.txt                             <- arguments used in this run
│   {run_title}_cenotetaker.log                   <- main log file
│
└───sequin_and_genome_maps/
│   │   {run_title}*gbf                           <- genome maps
│   │   {run_title}*fsa                           <- genome sequence
│   │   {run_title}*gtf                           <- feature table gtf format
│   │   {run_title}*tbl                           <- feature table sequin format
│   │   {run_title}*sqn                           <- non-human-readable sequin file for GenBank sub
│   │   {run_title}*cmt                           <- sequin comment file
│
└───ct_processing/
    │   --- many intermediate files ---

Ideas for downstream analyses

CheckV for virus genome completeness estimation.

BACPHLIP for phage lifestyle prediction (only use complete/near-complete phage genomes).

VContact3 for genome clustering and taxonomy.

iPHoP for prokaryotic virus host prediction.

Notes

Cenote-Taker 3 is under active development, so please open an issue if anything seems unusual or any errors occur. It's likely that I've not tested every parameter combination, and bugs will be a simple fix.

Citation

Cenote-Taker 3 for Fast and Accurate Virus Discovery and Annotation of the Virome.

Michael J. Tisza, Joseph F. Petrosino, Sara J. Javornik Cregeen

doi: https://doi.org/10.1101/2025.08.20.671380