Nov. 13, 2021

Downloading subsets of NCBI databases

Let's say you want to download all protein assigned to Clostridum from the non-redundant protein database at NCBI. We can achieve that by using common tools available in every Linux/Mac system out there.

This theoretical pipeline (we will be using pipes so let's call it a pipeline) looks like this:

stream/download file > decompress > parse records of interest > write to file

Now lets replace the actions by actual tools:


To stream the data we can use curl:

curl $URL


To decompress the stream of data from the previous command we can use gzip:

gzip -d -

parse records of interest and write to file

To parse and write the output we can write a quick python script that reads from stdin and write only the records that contains the string of interest in the record header:

import sys
copy_is_on = False

filter_string = sys.argv[1]

with open("output.fa", "w", buffering=100000) as fh:
    while True:
        line = sys.stdin.readline()
        if line.startswith(">"):
            if filter_string in line.lower():
                copy_is_on = True
                copy_is_on = False

        if copy_is_on:

Next we wrap everything into a single command:

curl | gzip -d - | python3 clostridium

Enjoyed this content? Don't forget to share!

comments powered by Disqus