1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
|
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
|
