For example, if you are writing a new fasta file parser, your colleagues could ask you whether you have tested it with specific fasta file cases – e.g. how does it handle files with an invalid header, or with blank lines within them.

By the way, this is more scientific, too :-)
Most scientists don’t know anything about programming, but they are better at the ‘testing an hypothesis’ work.

I have never used code reviews.
When I was working in a different group, where my group leader was a bioinformatician, he almost never looked at the code I wrote.

It is very difficult to have someone reviewing your scripts in bioinformatics. In fact, in the two years of experience that I have in this field, I have never seen anyone doing something like that.
It is really frustating when you are master student and don’t have much experience.

I think that pair programming and collective code ownership (http://extremeprogramming.org/rules/movepeople.html) could be a good solution to the problem, however the mentality of most of the bioinformaticians is far from understanding these issues.

Yes, code reviews work. At Smart Bear we’re admittedly biased – we sell an application that facilitates peer code review. But we have research and tips that are useful even *without* our software: http://smartbear.com/codecollab-white-paper.php