python-fmcs 1.0-1 / usr / share / doc / python-fmcs / TODO

Oh, there are so many ways to take this project. Almost all of them
require funding. Do you want to fund me?

Some possibilities include:

 o disconnected MCSes
 o user-defined atom typing
 o conversion to C++ (should be about 4x faster)
 o keep track of which canonical bondtypes are already in the
    subgraph. Prune the search if the count ever exceeds the maximum
    possible value.
 o support thresholds like "largest MCS found in at least 90%
    of the input structures."
 o use Python (long) integers as bitsets, rather than Python sets
    containing small integers. This might speed up the search.
 o list all subgraphs which have the same maximum size
 o the complete-rings-only implementation can do additional
    optimizations. Eg, if a fragment has no rings then remove
    all of its atoms which are ring atoms in the original structure.
 o option to report the MCS as a fragment SMILES instead of a SMARTS
 o use a crytpo hash algorithm to generate a universally
     unique SMARTS? Will that ever be useful?
 o improve selection so "1-10" doesn't require reading past the tenth
     structure.
 o add the ability to select based on identifiers
 o gzip input support