Parsing A Clustal Alignment File

Question

[s.charonis]

Hello BioStar Community,

I have a CLUSTAL alignment file with 500 protein sequences whose format is as such:

S1 DERY .....

S2 RKH .... ....

S500 HERKKK ....

where S1,S2,....S500 are Uniprot codes.

However, each of the 500 entries appears multiple times because each individual line is limited to n characters. So, S1,S2,...S500 appear multiple times and I need to parse each component together in order to get each individual sequence stored as a single string. The code for doing this on a single sequence is straightforward, first I put all Uniprot codes into a list so that I can identify each protein sequence and then I use a snippet like

S1 = '' 
for line in alignment: 
if line.startswith('3NY8A'):  
    line = line.lstrip('3NY8A') 
    line = line.rstrip('\n') 
    line = line.rstrip('\t')    
    templateSeq += line 
    templateSeq = ''.join(templateSeq.split())

But S1 is only one of the 500 sequences and I need to automate this procedure for all of the 500 sequences. What I think is necessary is to write a function (e.g. sequencetostring) and then implement this function over all lines in the alignment file. My implementation is

seq = ''
def sequence_to_string():
''' make each MSA sequence a string ''' 
    global seq 
    line = line.lstrip(unique_uniprot_id[i]) 
    line = line.rstrip('\n')
    line = line.rstrip('\t') 

    seq += line 
    seq = ''.join(seq.split())
    return seq

And then loop over the file with:

for i in range(len(unique_uniprot_id)):
  sequence_to_string()

But this does not return anything. Is there a way to modify my function so that it builds 500 strings (maybe I need to put them in a list?), one per protein sequence? Many thanks in advance!

Regards, Spyros

Answer 1

[Neilfws]

Any reason not to use the Bio.AlignIO module from Biopython to do this? It's generally better to use existing libraries than to reinvent the wheel.

[cl.parallel]

That would be my suggestion as well.

[Steve Moss]

Mine too! BioPython rocks!! ;) Do something like the following to convert from CLUSTAL to FASTA:

from Bio import AlignIO

input_file = open("uniprot_alignment.clustal", "rU")
output_file = open("uniprot_alignment.fasta", "w")

alignments = AlignIO.parse(input_file, "clustal")
AlignIO.write(alignments, output_file, "fasta")

output_file.close()
input_file.close()

Answer 2

[Damian Kao]

So you want to parse the alignment file and get each aligned sequence. As Neil suggested, I would just use BioPython.

For whatever reason, if you don't want to use BioPython, you can just do something like this:

seqs = {}
for line in alignmentFile:
   data = line.strip().split()
   seqID = data[0]
   seq = data[1]
   if not seqs.has_key(seqID):
      seqs[seqID] = seq
   else:
      seqs[seqID] += seq

for seqID, seq in seqs.items():
   print ">" + seqID 
   print ''.join(seq.split())

You don't need a separate function or anything like that. Use data structures to your advantage.

I am not entirely sure what your alignment file format looks like. Is it just a space delimited format? You might have to mess around with the parsing part. This should print out the sequences in fasta format.

Answer 3

[cl.parallel]

As a rule, you should try to avoid using global references. Your other references are confusing as well (where does line come from in sequencetostring?)

You could use a dictionary of lists:

def seqtostring(alignment):
  sequences = {}
  # header = alignment.readline() # if present
  for line in alignment: # the actual file, line-by-line
    if len(line): # spaces between groups
      name, seq = line.split('\t')
      if name not in sequences:
          sequences[name] = []
      sequences[name].append(seq.strip().upper())

  return sequences

with open(alignment_file) as openfile:
    seqs = seqtostring(openfile)
print id_of_interest, ''.join(seqs[id_of_interest])