parsing a fasta file using a generator ( python )

Question

I am trying to parse a large fasta file and I am encountering out of memory errors. Some suggestions to improve the data handling would be appreciated. Currently the program

Answer 1

A pyparsing parser for this format is only a few lines long. See the annotations in the following code:

data = """>1 (PB2) 
AATATATTCAATATGGAGAGAATAAAAGAACTAAGAGATCTAATGTCACAGTCTCGCACTCGCGAGATAC 
TCACCAAAACCACTGTGGACCACATGGCCATAATCAAAAAGTACACATCAGGAAGGCAAGAGAAGAACCC 
TGCACTCAGGATGAAGTGGATGATG 
>2 (PB1) 
AACCATTTGAATGGATGTCAATCCGACTTTACTTTTCTTGAAAGTTCCAGCGCAAAATGCCATAAGCACC 
ACATTTCCCTATACTGGAGACCCTCC"""

from pyparsing import Word, nums, QuotedString, Combine, OneOrMore

# define some basic forms
integer = Word(nums)
key = QuotedString("(", endQuoteChar=")")

# sequences are "words" made up of the characters A, G, C, and T
# we want to match one or more of them, and have the parser combine
# them into a single string (Combine by default requires all of its
# elements to be adjacent within the input string, but we want to allow
# for the intervening end of lines, so we add adjacent=False)
sequence = Combine(OneOrMore(Word("AGCT")), adjacent=False)

# define the overall pattern to scan for - attach results names
# to each matched element
seqEntry = ">" + integer("index") + key("key") + sequence("sequence")

for seq,s,e in seqEntry.scanString(data):
    # just dump out the matched data
    print seq.dump()
    # could also access fields as seq.index, seq.key and seq.sequence

Prints:

['>', '1', 'PB2', 'AATATATTCAATATGGAGAGAATAAAAGAACTAAGAGATCTAATGTCACAGTCTCGCACTCGCGAGATACTCACCAAAACCACTGTGGACCACATGGCCATAATCAAAAAGTACACATCAGGAAGGCAAGAGAAGAACCCTGCACTCAGGATGAAGTGGATGATG']
- index: 1
- key: PB2
- sequence: AATATATTCAATATGGAGAGAATAAAAGAACTAAGAGATCTAATGTCACAGTCTCGCACTCGCGAGATACTCACCAAAACCACTGTGGACCACATGGCCATAATCAAAAAGTACACATCAGGAAGGCAAGAGAAGAACCCTGCACTCAGGATGAAGTGGATGATG
['>', '2', 'PB1', 'AACCATTTGAATGGATGTCAATCCGACTTTACTTTTCTTGAAAGTTCCAGCGCAAAATGCCATAAGCACCACATTTCCCTATACTGGAGACCCTCC']
- index: 2
- key: PB1
- sequence: AACCATTTGAATGGATGTCAATCCGACTTTACTTTTCTTGAAAGTTCCAGCGCAAAATGCCATAAGCACCACATTTCCCTATACTGGAGACCCTCC

Answer 2

Without having a great understanding of what you are doing, I would have written the code like this:

def readFastaEntry( fp ):
    name = ""
    while True:
        line = name or f.readline()
        if not line:
            break
        seq = []
        while True:
            name = f.readline()
            if not name or name.startswith(">"):
                break
            else:
                seq.append(name)
        yield (line, "".join(seq))

This gathers up the data after a starting line up to the next starting line. Making seq an array means that you minimize the string joining until the last possible moment. Yielding a tuple makes more sense than a list.

Answer 3

Have you considered using BioPython. They have a sequence reader that can read fasta files. And if you are interested in coding one yourself, you can take a look at BioPython's code.

Edit: Code added

def read_fasta(fp):
    name, seq = None, []
    for line in fp:
        line = line.rstrip()
        if line.startswith(">"):
            if name: yield (name, ''.join(seq))
            name, seq = line, []
        else:
            seq.append(line)
    if name: yield (name, ''.join(seq))

with open('f.fasta') as fp:
    for name, seq in read_fasta(fp):
        print(name, seq)

Answer 4

def read_fasta(filename):
    name = None
    with open(filename) as file:
        for line in file:
            if line[0] == ">":
                if name:
                    yield (name, seq)
                name = line[1:-1].split("|")[0]
                seq = ""
            else:
                seq += line[:-1]
        yield (name, seq)