[Biojava-l] Google Summer of Code 2014 - Call for project ideas

Andreas Prlic andreas at sdsc.edu
Thu Feb 6 21:47:00 UTC 2014

Hi Amr,

During previous GSOC projects we  let the students build up new modules and
add major new functionality to the project. My preference would be to do
something similar again, if there is another round of GSOC. I would not
have time to contribute much this year, since at the UCSD side of things we
are getting an increasing number of students that are local. Having said
that, in principle there is no problem with letting a GSOC student work on
existing modules, but this will require much more planning and coordination
than is necessary when working on something new and independent.

On Thu, Feb 6, 2014 at 1:56 AM, Amr AL-HOSSARY <amr_alhossary at hotmail.com>wrote:

> I think we can start thinking in redesigning the structure module, getting
> rid of the current design, which is tightly coupled to PDB representation.

What is the problem with being close to PDB? This is the whole point of the
protein structure module... I would prefer to build on top of the current
data model, extending it and adding new features. (such as the support for
vecmath, that we discussed recently)

> Some of the details would include:
> - Separating I/O from the model (i.e. creating specialized I/O classes
> PDBReadr/Writer, MOL2Reader/Writer, etc).

This can be done on top of the current design. Since we already have
support for PDB/mmCif parsing, working on those parsers does not add much
new value to the project and I would discourage from advertising such a
project to students.

It would be great to have a Mol2Reader/writer. This is new functionality
that we currently don;t have.

> - redefining terms like group, molecule, bond, and model.

I guess I need more explanation why you would want to redefine those. The
current definitions are very close to the way PDB represents those.

> As well as
> redistributing attributes among them. (e.g. a model is an entity that has
> properties, not just a list of chains.

Should be possible with some minor refactoring of the model. What
properties do you want to associate to models?

> Also an atom has more properties like
> partial charge. a carbon atom can be SP2 or SP3).

Also, that should be possible with some additions to the current design, no
need for a major rewrite.

> If the idea is accepted in principle, we can start discuss it here or
> off-list.

I prefer such discussions on the -dev list. We should encourage development
out in the open and this allows people to chime in.


More information about the Biojava-l mailing list