Not python, but maybe awk is also good.

The logic behind this is, that the first lines have fields delimited by :. So there are at least 2 fields. The sequence itself have no :, so the field number is 1.

$ awk -v FS=":" '$1=="ACCESSION" {gsub(" ", "", $2); print ">"$2} NF==1 {print}' a.txt
>U49845
GATCCTCCATATACAACGGTATCTCCACCTCAGGTTTAGATCTCAACAACGGAACCATTGCCGACATGAG
ACAGTTAGGTATCGTCGAGAGTTACAAGCTAAAACGAGCAGTAGTCAGCTCTGCATCTGAAGCCGCTGAA

The gsub is used for removing white spaces.

fin swimmer

EDIT:

Here the same logic in python:

import sys

with open(sys.argv[1], "r") as genefile:
    for line in genefile:
        fields = line.split(":")
        if(fields[0] == "ACCESSION"):
            print(">"+fields[1].strip())
        if(len(fields) == 1):
            print(fields[0].strip())

Run it with python extract.py a.txt

I am sure there are more efficient and concise way of doing this, but this gets you the desired result.

#!/usr/bin/Python

# file to write the results
fasta = open("resultingFasta.fa","a")

# list to capture results
result = []

with open("dna.txt") as fh:
    for line in fh:
        if(line.startswith(("LOCUS","ORGANISM","AUTHORS","TITLE","JOURNAL","PUBMED","SOURCE")) ):
               # skip for lines starting with these
               continue 

        elif(line.startswith("ACCESSION")):
              # capture accession ID
              line = line.rstrip() # skip any trailing characters
              acc = line.split(": ")    # splitting ACCESSION: U49845
              result.append(">"+acc[1])# append accession ID to a list

        else:
              # the line with nucleotide sequence
              line = line.rstrip()
              result.append(line)   # append result to a list


for x in result:
    fasta.write(x+"\n") # writing it out to a file