I would strongly recommend to use Python and Biopython. In Biopython, there are some libraries to manipulate alignment, count sites, assign position. In a few lines, you could be able to do what you want. And even more.

You can do this with a simple selections, depending on your platform the code will vary. 

Here are the steps I would take.

1st: renumber all amino acids based on the alignment, to define a parameter that establishes an easy to use residue by residue correspondence. 

2nd: select all sequences that do not contain an arginine at 174, or any residue you do not want. Your software should have documentation for it's particular selection syntax which should use logical operators such as "and", "not",  "or" etc 

3rd: delete selected sequences

I would strongly recommend to use Python and Biopython. In Biopython, there are some libraries to manipulate alignment, count sites, assign position. In a few lines, you could be able to do what you want. And even more.

I strongly recommend to use a simple code in Perl, Python (even c++), too.

In this way, you do not face to any problem in order to manipulate big data.

Perl will do it nicely. Here follows a possible logic.

#1 get the aln into a hash, %ALN  (seqname => sequence)

$x = n; # n = position in aln of residue of interes.

$reswanted = 'X'; # where X = (A .. Y) residue you want to be present.

foreach $seqname (keys %ALN){

     my $sequence = $ALN{$seqname};

     my $res = substr ($sequence, $x-1,1);

     if ($res neq $reswanted  ){

            delete  $ALN{$seqname};

     }

}

## Then you can print the hash

Thank you all for the replies.

I used biopython to solve my problem.  Here is the python program I wrote (it's my first attempt at python so any critiques are welcome):

from Bio import AlignIO

from Bio.Align import MultipleSeqAlignment

alignment = AlignIO.read("MyProt.fasta", "fasta")      

goodseqs1 = MultipleSeqAlignment([])  

goodseqs2 = MultipleSeqAlignment([])  

badseqs = MultipleSeqAlignment([])     

for sequence in alignment:

    if sequence.seq[48] == "F":

        goodseqs1.append(sequence)  

    elif sequence.seq[48] == "Y":

        goodseqs1.append(sequence) 

    else:

        badseqs.append(sequence)      

for sequence in goodseqs1:     

    if sequence.seq[46] == "Q":

        goodseqs2.append(sequence)

    elif sequence.seq[46] == "R":

        goodseqs2.append(sequence)

    elif sequence.seq[46] == "G":

        goodseqs2.append(sequence)

    elif sequence.seq[46] == "I":

        goodseqs2.append(sequence)

    else:

        badseqs.append(sequence)

AlignIO.write(goodseqs2, "SCREENED_SEQS.FASTA", "fasta") 

AlignIO.write(badseqs, "Discard.fasta", "fasta")  

print("Alignment length %i" % alignment.get_alignment_length())  

print("initial # of seqs", len(alignment)) 

print("seqs passing first screen:", len(goodseqs1))   

print("final # of seqs", len(goodseqs2))  

print("output files are: SCREENED_SEQS.FASTA and Discard.fasta") 

You can do in one screen I think (I didn't try):

from Bio import AlignIO

from Bio.Align import MultipleSeqAlignment

alignment = AlignIO.read("MyProt.fasta", "fasta")

goodseqs = MultipleSeqAlignment([])

badseqs = MultipleSeqAlignment([])

for sequence in alignment:

     if sequence.seq[46] in ["Q","R","G","I"] and sequence.seq[48] in ["F","Y"] :

        goodseqs.append(sequence)

     else:

        badseqs.append(sequence)

AlignIO.write(goodseqs, "SCREENED_SEQS.FASTA", "fasta")

AlignIO.write(badseqs, "Discard.fasta", "fasta")

print("Alignment length %i" % alignment.get_alignment_length())

print("initial # of seqs", len(alignment))

print("final # of seqs", len(goodseqs))

print("output files are: SCREENED_SEQS.FASTA and Discard.fasta")

Yes you are correct Romain!  Your version worked perfectly.  Thank you very much for your help.

https://github.com/dukecomeback/bad-sequence-remover