RE: [Blue-obelisk] Re: Python Cheminformatics Framework

[Warren D. <wa...@de...>]

> I think that using the currently available cheminformatics 
> toolkits from Python is a much better idea than 
> reimplementing algorithms in a different language. For Java 
> libraries, you have Jython; for Openbabel, you have a Python 
> interface (and some nice glue too, if you use Pybel). 
> Implementing algorithms correctly is very tricky, and it's 
> better to take advantage of the large user bases of OB and 
> CDK (and their very active development communities) rather 
> than attempting to reimplement the wheel

I disagree.  OELib/OpenBabel is encumbered by the GPL, making it legally impractical for use as a core library in non-GPL projects, and CDK is only suitable for use in situations where one can accept the memory and performance overheads of a JVM.

To my knowledge, there is no clean, lightweight, and language-neutral core cheminformatics library yet to be found as open-source under a non-viral license that would mesh well with Python.  This is still huge unmet need for developers of standalone cheminformatics tools, and we should not discourage anyone from pursuing this goal with vigor.

However, CDK and the associated efforts are so well established that for in-house systems or academic products, it would indeed be far better to accept the overhead of a JVM and leverage CDK than to go off and write your own C or C++ library from scratch.  Likewise, if you're not planning on ever creating non-GPL products from your code, then OpenBabel would make a fine choice as well.

In summary, although the needs of the "Java world" are well met by CDK, and the needs of the "FSF socialist utopia" are well met by OpenBabel, the needs of traditional application developers for an unresricted open-source ".so" or ".DLL" that does chemistry well, and that would fit best with C/Python, remain unmet.

Cheers,
Warren