Run Python under strace sometime. Laugh as it attempts to open thousands of non-existent files just to start up. Cry when you imagine that happening on a thousand-node scientific computing cluster with an NFS root. IIRC we saw something on the order of 10-20 minutes to get Python started on every node of a Blue Gene allocation.

Ditto for Perl because of BioPerl (used in the Human Genome Project) & PDL (which IIRC was being used somewhere at NASA).

links: http://en.wikipedia.org/wiki/BioPerl | http://pdl.perl.org/

I believe Perl in bioinformatics was a combination of three or four factors. At that time, sequence analysis was mostly equivalent to string processing and executing command-line programs. Perl was a natural fit. Bioinformatics work used Suns running Sybase, so the researchers knew the Unix way of doing things, which is what perl builds on. This was the early days of the web, and bioinformatics people were very much into data sharing. They used perl to write CGI programs (both cgi-lib.pl and CGI.pm were written by bioinformatics people), so they had one language which they could use for sequence research, for system administration, and for data sharing.

Other languages evolved through a different path. Chemistry, for example, tends to be more Python related. (I develop software for that field.) I think it's because the chemical graph data structure is harder to write naturally in Perl. Gene expression analysis uses a lot of R.