How to remove sequence reads contains more than 2 X from multifasta file?

Question

I have 5000 protein sequences in one multifasta file. I found more reads have gaps as X in their reads. So, want to eliminate those reads completely (Whole protein seq) from the file. I am keeping filter criteria as if a read contains morethan 2 X ( continuesly or anywhere in the read) should be removed. Thanks in advance for your help.

The input sequence looks like this

>Prot1 
ANSTVKKKKLLLYYYSSSEERXFGHYFGHYFGHFYVHFGFYVHCEDYHF
>Prot2
ANSTVKKKKLLLYYYSSSEERXXXXXXXXXXXFGHYFGHYFGHFYVHFGFYVHCEDYHF
>Prot3
ANSTVKKKKLLLYYYSSSEERFGHYFGHYFGHFYVHFGFYVHCEDYHF

I want output Like this

>Prot3
ANSTVKKKKLLLYYYSSSEERFGHYFGHYFGHFYVHFGFYVHCEDYHF

Answer 1

A Biopytnon solution that works for .gz files too :-). The script make a script filter_faa.py:

#!/usr/bin/env python3

import sys
from Bio import SeqIO
from mimetypes import guess_type
import gzip

filename = sys.argv[1]

encoding = guess_type(filename)[1]
_open = partial(gzip.open, mode='rt') if encoding == 'gzip' else open

with _open(filename) as ffile:
    for seq_record in SeqIO.parse(ffile, 'fasta'):
        if seq_record.seq.count('X') < 3:
            print(seq_record.format("fasta"))

Then usage is

python3 filter_faa.py my.faa > my_filtered.faa 

The my_filtered.faa then will contain

>Prot1
ANSTVKKKKLLLYYYSSSEERXFGHYFGHYFGHFYVHFGFYVHCEDYHF

>Prot3
ANSTVKKKKLLLYYYSSSEERFGHYFGHYFGHFYVHFGFYVHCEDYHF

The filtering condition is hard-coded. You could potentially parametrise it if you feel like later you might want to filter by a different number of unknown aa.

Answer 2

Updated: as pointed out, this will not work for a multifasta as asked for in your original question. However it would work fine for a single line fasta

---

You could accomplish this with a single line of bash code if you would like (this assumes your fasta is in gzip format, if not just use cat):

zcat prot.fasta.gz | grep -E -v -B1 '(^>.*|^(.*X.*){3}$)' | grep -E -v '^--$'  - | gzip > prot.filtered.fasta.gzip

The -v option of grep inverts the regex, and the -B1 will print out the line occuring before the current context. The -B1 introduces -- between the matches, which is why the second grep is there. Then you can immediately pipe it back into a zipped file