ensembl_ecoli.zsh

#!/usr/bin/zsh

wget "ftp://ftp.ensemblgenomes.org/pub/bacteria/release-36/species_EnsemblBacteria.txt"

mv species_EnsemblBacteria.txt species.txt

# |head -n 2
ECOLI=("${(@f)$(grep 'Escherichia coli' species.txt | cut -f 2,5)}")

for line in $ECOLI; do
	NAME=$(echo "$line" | cut -f 1)
	ASM=$(echo "$line" | cut -f 2 | tr ' #/:' '____')
	
	FILE_NAME="E${NAME#e}.$ASM.dna.toplevel.fa.gz"

	if [[ ! -e "$FILE_NAME" ]]; then
		FILE_URL=$(curl -s -L -o- "http://bacteria.ensembl.org/$NAME/Info/Index/" |
			grep ftp |
			grep -oPe "ftp://ftp.ensemblgenomes.org/pub/bacteria/release-36/fasta/[^\"]*" |
			head -n 1 |
			paste -d '' - <(echo "E${NAME#e}.$ASM.dna.toplevel.fa.gz"))

		echo "$NAME \t$FILE_URL"
		wget -nv "$FILE_URL"
	fi
done

## broken species:
# escherichia_coli_gca_001499595
# escherichia_coli_e1728
# escherichia_coli_k_12_gca_000981485
# escherichia_coli_o26_h11


# deal with synthetic explicitly
wget -nv "ftp://ftp.ensemblgenomes.org/pub/bacteria/release-36/fasta/bacteria_50_collection/synthetic_escherichia_coli_c321_deltaa/dna/Synthetic_escherichia_coli_c321_deltaa.ASM47403v1.dna.toplevel.fa.gz"