Gene_fetch

This tool fetches gene sequences from NCBI databases based on taxonomy IDs (taxids) or taxonomic information. It can retrieve both protein and nucleotide sequences for various genes, including protein-coding genes (e.g., cox1, cytb, rbcl, matk) and rRNA genes (e.g., 16S, 18S).

Feature highlight

• Fetch protein and/or nucleotide sequences from NCBI GenBank database.
• Handles both direct nucleotide sequences and protein-linked nucleotide searches (CDS extraction includes fallback mechanisms for atypical annotation formats). .
• Support for both protein-coding and rDNA genes.
• Single-taxid mode (-s/--single) for retrieving a specified number of target sequences for a particular taxon (default length thresholds are reduced (protein: 50aa, nucleotide: 100bp)).
• Customisable length filtering thresholds for protein and nucleotide sequences.
• Automatic taxonomy traversal: Uses fetched NCBI taxonomic lineage for a given taxid when sequences are not found at the input taxonomic level. I.e., Search at given taxid level (e.g., species), if no sequences are found, escalate species->phylum until a suitable sequence is found.
• Validates fetched sequence using higher taxonomy, avoiding potential taxonomic homonyms.
• Robust error handling, error and progress logging, and NCBI API rate limits (10 requests/second).
• Handles complex sequence features (e.g., complement strands, joined sequences, WGS entries) in addition to 'simple' cds extaction (if --type nucleotide/both). The tool avoids "unverified" sequences and WGS entries not containing sequence data (i.e. master records).
• 'Checkpointing': if a run fails/crashes, the script can be rerun using the same arguments and it will resume from where it stopped.
• When more than 50 matching sequences are found for a sample, the tool fetches summary information for all matches (using NCBI esummary API), orders them by length, and processes the top 10 longest sequences.

Contents

• Installation
• Usage
• Examples
• Input
• Output
• Cluster
• Supported targets
• Notes
• Benchmarking
• Future developments
• Contributions and citation

Installation

First, clone the Gene Fetch GitHub repository to your current path, and enter the Gene Fetch installation directory

git clone https://github.com/bge-barcoding/gene_fetch

cd gene_fetch

Run the commands below to install the necessary dependencies and activate the Conda environment. Conda must be installed.

conda env create -n fetch -f fetch.yaml

conda activate fetch

Alternatively, you can install the dependencies below directly or in your own Conda environment

conda install python>=3.9 pip
pip install ratelimit>=2.2.1
pip install biopython>=1.80

Usage

python gene_fetch.py -g/--gene  --type  -i/--in  -o/--out  

• --h/--help: Show help and exit.

Required arguments

• -g/--gene: Name of gene to search for in NCBI GenBank database (e.g., cox1/16s/rbcl).
• --type: Sequence type to fetch; 'protein', 'nucleotide', or 'both' ('both' will initially search and fetch a protein sequence, and then fetches the corresponding nucleotide CDS for that protein sequence).
• -i/--in: Path to input CSV file containing sample IDs and TaxIDs (see Input section below).
• i2/--in2: Path to alternative input CSV file containing sample IDs and taxonomic information for each sample (see Input section below).
• o/--out: Path to output directory. The directory will be created if it does not exist.
• e/--email and -k/--api-key: Email address and associated API key for NCBI account. An NCBI account is required to run this tool (due to otherwise strict API limitations) - information on how to create an NCBI account and find your API key can be found here. ####= Optional arguments
• --protein_size: Minimum protein sequence length filter. Applicable to mode 'normal' and 'single-taxid' search modes (default: 500).
• --nucleotide_size: Minimum nucleotide sequence length filter. Applicable to mode 'normal' and 'single-taxid' search modes (default: 1500).
• s/--single: Taxonomic ID for 'single-taxid' sequence search mode (-i and -i2 ignored when run with -s mode). 'Single-taxid' mode will fetch all target gene or protein sequences on GenBank for a specific taxonomic ID.
• --max-sequences: Maximum number of sequences to fetch for a specific taxonomic ID (only applies when run in 'single-taxid' mode).

Examples

Fetch both protein and nucleotide sequences for COI with default sequence length thresholds.

python gene_fetch.py -e [email protected] -k your_api_key \
                    -g cox1 -o ./output_dir -i ./samples.csv \
                    --type both

Fetch rbcL nucleotide sequences using sample taxonomic information, applying a minimum nucleotide sequence length of 1000bp

python gene_fetch.py -e [email protected] -k your_api_key \
                    -g rbcl -o ./output_dir -i2 ./taxonomy.csv \
                    --type nucleotide --nucleotide_size 1000

Retrieve 1000 available matK protein sequences >400aa for Arabidopsis thaliana (taxid: 3702).

python gene_fetch.py -e [email protected] -k your_api_key \
                    -g matk -o ./output_dir -s 3702 \
                    --type protein --protein_size 400 --max-sequences 1000

Input

Example 'samples.csv' input file (-i/--in)

Example 'samples_taxonomy.csv' input file (-i2/--in2)

Output

'Normal' mode

output_dir/
├── nucleotide/                 # Nucleotide sequences. Only populated if '--type nucleotide/both' utilised.
│   ├── sample-1_dna.fasta   
│   ├── sample-2_dna.fasta
│   └── ...
├── sample-1.fasta              # Protein sequences.
├── sample-2.fasta
├── sequence_references.csv     # Sequence metadata.
├── failed_searches.csv         # Failed search attempts (if any).
└── gene_fetch.log              # Log.

sequence_references.csv output example

'Single-taxid' mode

output_dir/
├── nucleotide/                      # Nucleotide sequences. Only populated if '--type nucleotide/both' utilised.
│   ├── ACCESSION1_dna.fasta   
│   ├── ACCESSION2_dna.fasta
│   └── ...
├── ACCESSION1.fasta                 # Protein sequences.
├── ACCESSION2.fasta
├── fetched_nucleotide_sequences.csv # Only populated if '--type nucleotide/both' utilised. Sequence metadata.
├── fetched_protein_sequences.csv    # Only populated if '--type protein/both' utilised. Sequence metadata.
├── failed_searches.csv              # Failed search attempts (if any).
└── gene_fetch.log                   # Log.

fetched_protein|nucleotide_sequences.csv output example

Running gene_fetch on a cluster

• See '1_gene_fetch.sh' for running gene_fetch.py on a HPC cluster (SLURM job schedular).
• Edit 'mem' and/or 'cpus-per-task' to set memory and CPU/threads allocation.
• Change paths and variables as needed.
• Run '1_gene_fetch.sh' with:

sbatch 1_gene_fetch.sh

Supported targets

Gene Fetch does function with other targets than those listed below, but it has hard-coded name variations and 'smarter' searching for the below targets. More targets can be added into script (see 'class config').

• cox1/COI/cytochrome c oxidase subunit I
• cox2/COII/cytochrome c oxidase subunit II
• cox3/COIIIcytochrome c oxidase subunit III
• cytb/cob/cytochrome b
• nd1/NAD1/NADH dehydrogenase subunit 1
• nd2/NAD2/NADH dehydrogenase subunit 2
• rbcL/RuBisCO/ribulose-1,5-bisphosphate carboxylase/oxygenase large subunit
• matK/maturase K/maturase type II intron splicing factor
• 16S ribosomal RNA/16s
• SSU/18s
• LSU/28s
• 12S ribosomal RNA/12s
• ITS (ITS1-5.8S-ITS2)
• ITS1/internal transcribed spacer 1
• ITS2/internal transcribed spacer 2
• tRNA-Leucine/trnL

Benchmarking

Future Development

• Add optional alignment of retrieved sequences
• Add support for direct GenBank submission format output
• Enhance LRU caching for taxonomy lookups to reduce API calls
• Further improve efficiency of record searching and selecting the longest sequence
• Add support for additional genetic markers beyond the currently supported set

Contributions and citations

GeneFetch was written by Dan Parsons & Ben Price @ NHMUK (2024).

If you use GeneFetch, please cite our publication: XYZ()

If you have any questions or suggested improvements, please do get in touch in the issues!