VirBot: an RNA viral contig detector for metagenomic data

Without relying on cultivation, metagenomic sequencing greatly accelerated the novel RNA virus detection. However, it is not trivial to accurately identify RNA viral contigs from a mixture of species. The low content of RNA viruses in metagenomic data requires a highly specific detector, while new RNA viruses can exhibit high genetic diversity, posing a challenge for alignment-based tools.

Here, we develop VirBot, an easy-to-use yet effective RNA virus detection tool from metagenomic data. It takes assembled contigs as input and detect ones from RNA viruses. For each detected contig, we assign it the likely taxon label according to the encoded protein that has the highest score against our pHMMs.

VirBot is deisgned based on a comprehensive RNA viral protein family database. Besides the hallmar gene in RNA virus, RdRps, we also leverage other essential proteins, including capsid proteins, envelop proteins, viral auxiliary proteins, etc. By using the adaptative bit score cutoff, VirBot shows its higher specificity in metagenomic dataset and sensitivity in novel RNA virus dataset. VirBot supports identifying contigs as short as 500bp. The construction of RNA viral pHMMs and key components of VirBot are shown as Fig (a,b). For each RNA viral pHMM, we use voting strategy among the clustered viral proteins to determine the taxon, which is used for the taxonomic assignment of the detected contigs.

We validated VirBot in various scenarios and benchmarked it with other 7 RNA virus detection tools. VirBot achieves higher recall than other tools. And it also demonstrates high specificity in metagenomic data that only contains a small number of RNA viruses. Here, we briefly show the result of VirBot on metagenomics data (Fig (c)) and pure RNA virus dataset (Fig (d)).

(a) Construction of the RNA viral pHMMs database. (b) Sketch of the key components of VirBot. (c) Detection performance on simulated data: ERR1992810 (left) and ERR2185279 (right). (d) Recall on RNA viral datasets: RNA phages dataset (left) and marine water RNA virome dataset (right).
where the metagenomics dataset contains sequences from 82 eukaryotes, 365 prokaryotes, and DNA/RNA viruses; and for the RNA virus datasets, one comprises 8,849 RNA phages that were barely detected before, while eukaryotic viruses dominate known RNA viruses; another dataset is an RNA virome sample sequenced from marine water containing 114,139 RNA viral seqeunces.

Dependency:

• Prodigal 2.6.3
• HMMER3 3.3.2
• DIAMOND 2.0.15
• pandas 1.5.2
• python 3.x

Quick install

We highly recommend using conda or mamba to install all the dependencies. To install, please download VirBot by "git clone"

git clone https://github.com/GreyGuoweiChen/RNA_virus_detector
cd RNA_virus_detector

# create the environment and install the dependencies using conda or mamba
mamba env create -f environment.yml

# activate the environment
conda activate virbot

# install
pip install .

If you are unable to install GIT-LFS, you can instead manually download the references file from OneDrive and place it directly into the virbot/data directory before running the pip install step.

Usage:

The VirBot is friendly to use. It requires input as fasta format and will return the fasta format output file with the identified RNA virus sequences. If you have not installed using the pip install command above you will need to add virbot/VirBot.py to your path or call the script directly.

# If the environment variables path is correctly set and you run VirBot.py as an executable file.
virbot [--input INPUT_CONTIG] [--output OUTPUT_DIRECTORY] [--sen] [--threads]

# If you run VirBot.py as python script.
python VirBot.py [--input INPUT_CONTIG] [--output OUTPUT_DIRECTORY] [--sen] [--threads]

Options

--input: The input contig file in fasta format.
--output: The output directory (default: VB_result).
--sen (Optional): Use the sensitive mode for VirBot.
--threads (Optional): The threads number run for HMMER and DIAMOND (default: 8)

Example:

virbot --input test/test_input.fa

virbot --input test/test_input.fa --output VB_result --sen --threads 8

Benchmark datasets:

Please check the link in OneDrive. This file is the same as the GIT-LFS hosted file virbot/data/ref.zip containing the models required by virbot. In addition we provide sequence files which can be found in virbot/data/seq.zip