As the co-author of FragIt(web, code), a piece of
software written in Python using the Open Babel API, to help setup
and fragment large molecules in fragment based methods, I have to
return to this (excellent) piece of software to make it work with
Molecular Fragmentation with Conjugate Caps (MFCC) methods which is
the approach that my new work-place has taken as their method of
choice in fragment based methods. Currently, FragIt only (officially)
supports the Fragment Molecular Orbital (FMO) method.
Step I of this transformation is to
realize that FMO and MFCC are two very different beasts. While the
theory of FMO is more hairy, the information that FMO needs to run is
vastly less than MFCC. In FMO, you specify pairs of atoms between
which you wish to fragment and then everything goes along nicely
whereas in MFCC you build fragments, attach caps and build extra
fragments from those caps (called conjugate caps). The latter is very
tough to do generally which is the approach and idea of FragIt in the
first place: Mostly because getting the correct SMARTS you need is very
cumbersome. As with the previous approach of FragIt, I'll make it work for proteins by selecting the appropriate SMARTS and program it in such a way that extensions to other systems would be straight-forward by simply figuring out additional patterns.
Right now my approach is to build
“capping”-SMARTS that select atoms around places of fragmentation
and build those caps in Open Babel. Non-trivial task is non-trivial I
must admit.
Stay tuned as I explore the SMARTS and
code needed to accomplish this.
Seems like one could solve a lot of problems with sensible defaults for caps and conjugate caps. But, yes, a completely general implementation will be a major change.
ReplyDeleteHi, we have implemented MFCC in our scripting framework PyADF, also using Openbabel and SMARTS. Maybe you want to have a look at that or can reuse some of our code? For more on PyADF, see www.pyadf.org; we still didnt manage to put together a release version, but if you are interested I can send you the current version of the code.
ReplyDeleteHi Christoph,
DeleteThis sounds interesting indeed. I will write you at some point to have a look, your API looks very interesting :)
cheers