VirusTaxo: Taxonomic classification of viruses from metagenomic contigs

VirusTaxo has an average accuracy of 93% at genus level across DNA and RNA viruses.

1. Running VirusTaxo

Requirements

• python >= 3.8
• Linux

Installation

• Cloning the repository

  git clone https://github.com/omics-lab/VirusTaxo

• Creation of Python Virtual Environment

  cd VirusTaxo
  python3 -m venv environment
  source ./environment/bin/activate

• Installation of Python Packages

  pip install -r requirements.txt

2. Predict virus taxonomy from fasta file using prebuilt database

• Download prebuilt databse (database.v2_2024) of VirusTaxo

gdown "https://drive.google.com/uc?id=1gz0n5oHomWjpT0HXsrqh8hTLqmqiqgJs"

# Extract db files
tar -xvzf vt_db_jan21_2024.tar.gz

• Database files and usage:

• Assemble the metagenomic contigs from your metavirome or metagenomic library.

  • Perform de novo assembly using MEGAHIT:

  megahit -1 file_R1.fq -2 file_R2.fq --min-contig-len 500 -o contig.fasta

• Usage of predict.py:

python3 predict.py -h

usage: predict.py [-h] --model_path MODEL_PATH --seq SEQ [--output_csv OUTPUT_CSV] [--entropy ENTROPY] [--enrichment_score ENRICHMENT_SCORE]

options:
  -h, --help            show this help message and exit
  --model_path MODEL_PATH
                        Absolute or relative path of pre-built model
  --seq SEQ             Absolute or relative path of fasta sequence file
  --output_csv OUTPUT_CSV
                        Path to save the output CSV file (default: VirusTaxo_taxonomy_output.csv)
  --entropy ENTROPY     Entropy threshold for classification (default: 0.5)
  --enrichment_score ENRICHMENT_SCORE
                        Enrichment score threshold for classification (default: 0.8)

• Run with the sample contig.fasta file

python3 predict.py \
   --model_path /path/DNA_RNA_18451_k20.pkl \ # database file
   --seq ./Dataset/contig.fasta # query fasta file 

• Example output for 4 query sequences

Id              Length  Genus           Entropy Enrichment_Score
QuerySeq-1      219     Unclassified    1.000   0.000
QuerySeq-2      720     Betacoronavirus 0.000   0.973
QuerySeq-3      1540    Unknown         0.285   0.820
QuerySeq-4      1330    Lentivirus      0.000   0.987

3. Interpretation of output

• In the taxonomic rank column

  • NoHit: no k-mer overlap between the query and database.
  • Unknown: genus name is not assigned in the ICTV classification.
  • Unclassified: Entropy (default >= 0.5) or Enrichment_Score (default <= 0.8) is outside of cutoff.

• Lower Entropy (such as ≤=0.5) provides the higher level of prediction certainty. You can decrease Entropy cutoff for better prediction.

• Higher Enrichment_Score (such as >= 0.8) provides the higher level of prediction certainty. You can increase Enrichment_Score cutoff for better prediction. Enrichment_Score is the total number of k-mers mapped to the genera divided by total number of k-mers in the query sequence.

4. Build your custom database

• Preparing a metadata file in csv format. The metadata file must contain columns named Id and Genus. Example of metadata file is here:

• The sequnce IDs must match with the metadata IDs. Example of input fasta file is here.

  • Building database:

python3 build.py \
   --meta ./Dataset/RNA_meta.csv \ # provide your metadata file
   --seq ./Dataset/RNA_seq.fasta \ # provide your fasta file
   --k 17 \
   --saving_path /path/RNA.pkl

• Parameters

  • meta: Absolute or relative path of metadata file.
  • seq: Absolute or relative path of fasta sequence file.
  • k : It denotes the length of k-mer.
  • saving_path: Path to save a pickle file (A database File).

5. Method limitation and interpretation

• VirusTaxo is trained on known virus sequences and designed to predict taxonomy of virus sequences.

• Since VirusTaxo uses k-mer enrichment, non-virus sequences could be classified as virus due to random k-mer match. To use VirusTaxo, please make sure to remove non-virus sequences.

• If your sample contains non-virus sequences, it is highly recommeded to filter out non-viral sequences using tools like blast or DeepVirFinder dvf.py -i contig.fasta -o ./.

• To avoid contaimination with host sequences, please filter out those by mapping the reads to host reference genomes before using VirusTaxo.

6. Version history

7. Contact

Rashedul Islam, PhD (rashedul.gen@gmail.com)

8. Citation

Rajan Saha Raju, Abdullah Al Nahid, Preonath Chondrow Dev, Rashedul Islam. VirusTaxo: Taxonomic classification of viruses from the genome sequence using k-mer enrichment . Genomics, Volume 114, Issue 4, July 2022.