Fetching genomes from Ensembl

Downloads a human/mouse primary assembly and associated annotation in GTF format from Ensembl and checks for data integrity. Inspired from here.

fetch_genome.sh

#!/usr/bin/env bash

species="Human"
version=114

if [ "${species}" == "Human" ] || [ "${species}" == "human" ]; then
    species_short="homo_sapiens"
    species_long="Homo_sapiens.GRCh38"
elif [ "${species}" == "Mouse" ] || [ "${species}" == "mouse" ]; then
    species_short="mus_musculus"
    species_long="Mus_musculus.GRCm39"
else
    echo "The script works for human or mouse."
fi

# The first argument is the path to the file of interest; FASTA, GTF, ...
# The second argument is the path to the associated checksum file
fetch_check() {

    echo -e "\nDownloading ${1}..."
    wget -q "${1}"
    file="$(basename "${1}")"

    calculated_sum="$(sum "${file}")"
    echo "Calculated checksum: ${calculated_sum}"

    downloaded_sum="$(wget -q -O - "${2}" | grep "${file}" | cut -d " " -f1,2)"
    echo "Downloaded checksum: ${downloaded_sum}"

    if [ "${calculated_sum}" == "${downloaded_sum}" ]; then
        echo "Checksums match..."
    else
        echo "CHECKSUMS DO NOT MATCH!"
    fi

}

base_path="https://ftp.ensembl.org/pub/release-${version}"

fasta_path="${base_path}/fasta/${species_short}/dna/${species_long}.dna_sm.primary_assembly.fa.gz"
fasta_sum_path="${base_path}/fasta/${species_short}/dna/CHECKSUMS"

fetch_check "${fasta_path}" "${fasta_sum_path}"

gtf_path="${base_path}/gtf/${species_short}/${species_long}.${version}.gtf.gz"
gtf_sum_path="${base_path}/gtf/${species_short}/CHECKSUMS"

fetch_check "${gtf_path}" "${gtf_sum_path}"