[Biopython-dev] PopGen status and new developments

[Tiago Antão]

Hi,

On Wed, Nov 12, 2008 at 6:58 PM, Bruce Southey <bsouthey at gmail.com> wrote:
> I do have concerns about usefulness with regards to Biopython.

It is important to notice that having this application support has no
big impact on deployment of biopython. The only visible thing is some
tests reporting that the application doesn't exist. This is different
from adding a dependency on, say, scipy. I don't think that this
imposes any maintenance/installation hurdle at large. I think, this is
actually a non-problem on the deployment stage, at least.

> How widespread is the application?

The application is fairly new (genepop, on the other hand is widely
used and old). I cannot answer that question. I know of some people
using it, but it is my small, biased, universe. I would guess that
currently the number is small.

Is there a policy to only support widespread applications?

> What platforms is it released under (DOS only or some version of windows
> version like XP or Vista or Windows 7)?

There is a Dos and Windows frontend. I actually asked the code to the
authors and they gave me access to it. I have compiled a Linux
version, but I don't know if they are going to make it available.

> Is the application well supported and will it continue to be supported?

Regarding current support, I can subjectively say that the authors
answer my queries rather fast. Regarding the future, I dont know.

> Under what terms is the application 'free'?

Much software available in this field is made available without no
regards for licensing issues. This is already the case for the
supported Fdist application (source available, no license).
This is problem in the field, where people make things available
without much concern for licensing issues. Some people don't care that
much about that, they just "make things available".
So, if there is a policy to only support applications for which there
is a clear license, then this one is out (and some code has to be
removed from the current PopGen module, by the way). I never link the
code in, I just invoke it (these are mostly wrappers), so there should
be no legal issues in any case, I suspect.

There is a chicken and egg problem here that needs to be fought: In
population genetics there is no widespread tradition of making things
open (not because people want closed solutions, but mostly because
people don't think about these issues). There is also little tradition
in coding (people want ready made solutions. The coding people is
relatively few and mostly R based) than in other areas. As an example:
i don't know of many direct users of fdist code, but know lots of
people which use applications made on top of that code.

By the way, Simcoal is GPL (and there are more examples of open code
in population genetics, of course).

> How does this integrate into your ideas for Popgen?

Very well. I have this stated philosophy, from the beginning, of using
existing applications and not reinvent the wheel. That being said, I
agree that a core statistic implementation should be done (even if
there are alternatives). But, mostly, for now, what is available in
Bio.PopGen are intelligent wrappers.

> Would it work like say clustalw where you output something from Biopython,
> run the application and perhaps import something back into Biopython?

Yep, it accepts genepop files and the output is fully parsed back.
This is still not the case, by the way, with simcoal where the output
is not usable (arlequin is needed to analyze the results). I need to
do an arlequin parser, that would solve the problem.

> If the application requires major data formatting then you would have to

It doesn't require any formatting at all as the de facto standard
format in the area (genepop) is supported and the results are parsed
back.

Tiago