AlignIO gives 'AssertionError' when reading emboss alignment files - bioinformatics

Question

I have been stuck on a problem for three days... searched everywhere, posted on Biostar, still waiting for EMBL to respond to emails... would make a bounty if I had more rep.
After aligning sequences with EMBOSSwin needle() (pairwise global alignments) I get alignment files in pair format, with a .needle file extension. I want to use Biopython to read these alignments for later analysis.
I use AlignIO.read(open('alignment.needle'),'emboss') following the instructions in Biopython's AlignIO wiki but I keep getting an AssertionError.
My code:
>>> from Bio import AlignIO
>>> alignment = AlignIO.read(open("data/all/out/pair1_alignment.needle"), "emboss")
My error:
Traceback (most recent call last):
File "<console>", line 1, in <module>
File "C:\Python27\lib\Bio\AlignIO\__init__.py", line 423, in read
first = next(iterator)
File "C:\Python27\lib\Bio\AlignIO\__init__.py", line 370, in parse
for a in i:
File "C:\Python27\lib\Bio\AlignIO\EmbossIO.py", line 150, in __next__
assert seq.replace("-", "") != ""
AssertionError
Example Alignment File:
Download the alignment file here
Versions:
Windows 7
Python version 2.7.3
Biopython version 1.63
EMBOSS version 2.10.0-0.8
Clues:
I suspect this may be related to a warning message I kept getting when actually making the alignments, which was outputted by EMBOSS needle() function:
Warning: Sequence character string not found in ajSeqCvtKS

Answer 1

Duplicate post on BioStars, http://www.biostars.org/p/87226/#87399
This appears to be down to a subtle change in the EMBOSS output. You have an extremely old version, EMBOSS version 2.10.0 (February 2005), and your output file has lines like this:
gag 1288 -------------------------------------------------- 1287
Using a newer version of EMBOSS (e.g. 6.3.0), gives lines like this:
gag 1287 -------------------------------------------------- 1287
The Biopython parser is expecting the latter for alignment sections with no letters (e.g. when one sequence is much longer than the other), where the start and end coordinates agree. Please update your copy of EMBOSS, and then the parser should be happy. The current EMBOSS release is version 6.5.0.

Answer 2

The problem is that you're passing the wrong format file to Biopython. An explanation follows.
Formatting
The format of the file you've linked to is srspair (see the header of pair1_aligned.fasta). It's worth noting that this is not the FASTA format - that's an entirely different format.
Delving into the source of Biopython's EmbossIO, we can see that the EmbossIterator (which is called by AlignIO.read when the format is 'emboss') is only meant to handle the formats pair and simple (see Alignment formats for an explanation of the various formats).
Solution
If you export EMBOSS's output in the pair format (then call AlignIO.read as you have before), that should solve your problem.