mxk62 / pca Goto Github PK

Currently, the list of recognized functional groups is stored in an external file. Thus it can easily be misplaced or go out-of-sync between different coders.

It would be better to place them as a collection in the mongo database. Then, we would have a single, well-defined and definite resource.

Before the method get_group_descriptor() of a Reaction class is called the functional groups have to be initialized somewhere outside (currently, in rxnpca.py).

It would be much more 'object-oriented' approach if moved to the class itself.

Some of the molecular descriptors in Chemical class are lacking a docstring explaining their meaning. They should be added for future reference.

If explanation will benefit from including descriptor's mathematical formula, encode it in LaTeX!

Construction of the scoring function is based on an assumption that published reactions are subset of all possible reactions generated by the transforms. Currently it is violated as all available published reaction are included to the dataset; even ones not recreated by transforms.