17:18:36drmeisterBike: I've been thinking about the fep-benchmarks - we have a collection of molecules like those in that picture that I posted where every pair of molecules shares a substructure with a few differences.
17:18:57drmeisterWe need to identify the common part and the differences for any particular pair.
17:19:32drmeisterI exposed this boost::graph function: https://www.boost.org/doc/libs/1_54_0/libs/graph/doc/mcgregor_common_subgraphs.html
17:20:13drmeisterBut it's useless - the time the algorithm takes goes up really fast with the number of atoms. It can't handle more than 14 despite my attempts to improve the situation.
17:20:46drmeisterI removed hydrogens from the graphs - that cuts them down by almost half.
17:21:30drmeisterThen I added a criterion where the vertices being compared need to be closer to each other in 3D space (atoms have positions and common atoms are very close to each other in space) - but that didn't help.