a small piece of code to easily hack into GTF files. Some features not optimized for large genomes so may hog on memory

GTFparser.py

import sys
from itertools import groupby

def read_fasta(fasta_name):
    fh = open(fasta_name)
    faiter = (f[1] for f in groupby(fh, lambda line: line[0] == ">"))
    for head in faiter:
        head = head.next()[1:].strip()
        s = "".join(s.strip() for s in faiter.next())
        yield head, s

class GenericFeature(object):
	def __init__(self, name, uid, chrom, start, end, strand, parent, feat_type, misc):
		self.chrom = chrom
		self.start = start
		self.end = end
		self.strand = strand
		self.parent = parent
		self.children = []
		self.type = feat_type
		self.name = name
		self.id = uid
		self.info = misc

	def set_child(self, child_id):
		self.children.append(child_id)


class GTF(object):
	def __init__(self, gtf_file):
		self.file = gtf_file
		self.genes = {}
		self.transcripts = {}
		self.exons = {}
		self.summary = {}
		self.file_parser()
		self.make_chilren()

	def lazy_reader(self):
		with open(self.file) as h:
			for l in h:
				if l[0] != "#" and len(l) > 20:
					c = l.rstrip('\n').split('\t')
					yield c

	def file_parser(self):
		gtf_stream = self.lazy_reader()
		for row in gtf_stream:
			anno = {x.split(' ')[0]: x.split(' ')[1].rstrip(';').strip('"') for x in row[-1].split('; ')}
			if row[2] == "gene":
				if 'gene_name' not in anno:
					anno['gene_name'] = None
				feature = GenericFeature(anno['gene_name'], anno['gene_id'], row[0], row[3], row[4],
										 row[6], None, 'gene', {'gene_biotype': anno['gene_biotype']})
				self.genes[anno['gene_id']] = feature
			elif row[2] == 'transcript':
				if 'transcript_name' not in anno:
					anno['transcript_name'] = None
				feature = GenericFeature(anno['transcript_name'], anno['transcript_id'], row[0],
										 row[3], row[4], row[6], anno['gene_id'], 'transcript',
										 {'transcript_biotype': anno['transcript_biotype']})
				self.transcripts[anno['transcript_id']] = feature
			elif row[2] == 'exon':
				feature = GenericFeature(None, anno['exon_id'], row[0],
										 row[3], row[4], row[6], anno['transcript_id'], 'exon',
										 {'exon_number': anno['exon_number']})
				self.exons[anno['exon_id']] = feature

	def summarize(self):
		if self.summary == {}:
			self.summary = {
				'genes': len(self.genes),
				'transcripts': len(self.transcripts),
				'exons': len(self.exons)
			}
		print self.summary

	def make_chilren(self):
		for exon in self.exons.values():
			self.transcripts[exon.parent].children.append(exon.id)
		for transcript in self.transcripts.values():
			self.genes[transcript.parent].children.append(transcript.id)

	def make_transcriptome_json(self):
		chrom_wise = {}
		for gene in self.genes.values():
			for transcript_id in gene.children:
				transcript = self.transcripts[transcript_id]
				coords = []
				for exon_id in transcript.children:
					exon = self.exons[exon_id]
					coords.append([int(exon.start), int(exon.end)])
				if transcript.chrom not in chrom_wise:
					chrom_wise[transcript.chrom] = []
				chrom_wise[transcript.chrom].append({
					'transcript_id': transcript.id,
					'transcript.name': transcript.name,
					'gene.id': gene.id,
					'gene.name': gene.name,
					'seq_coords': coords,
					'strand': transcript.strand
				})
		return chrom_wise

	def make_genome_dict(self, fasta_file):
		genome_dict = {}
		for h,s in read_fasta(fasta_file):
			genome_dict[h] = s
		return genome_dict

	def make_transcriptome_fasta(self, genome_file, out_fasta):
		OUT = open(out_fasta, 'w')
		genome_dict = self.make_genome_dict(genome_file)
		chrom_wise_info = self.make_transcriptome_json()		
		for chrom in chrom_wise_info:
			seq = genome_dict[chrom]
			for t in chrom_wise_info[chrom]:
				t_seq = []
				for coord in t['seq_coords']:
					t_seq.append(seq[coord[0]:coord[1]])
				OUT.write(">%s\n%s\n" % (t['transcript_id'], "".join(t_seq)))
		OUT.close()
		return True
				
if __name__ == "__main__":
	gtf_file = sys.argv[1]
	gtf = GTF(gtf_file)
	gtf.summarize()
	gtf.make_transcriptome_fasta('./resource/genome.fa', './resource/transcripts.fasta')