[Open Babel] New Fingerprint Types

[Chris M. <c.m...@ga...>]

ern...@ba... wrote:
>> I'm currently experimenting with a modified FP3 fingerprint that uses 
>> multiple bits to code not only the existence of a pattern but also it's 
>> count.
>>

Craig A. James wrote:
> I've thought about something like this before. 
> Another thing to consider: The distribution of counts for functional groups
 > is wildly different.  For example, the difference between three and 
four
 > nitro groups might be very significant, but the difference between 
three
 > or four methyls is not.  Similarly, the presence of a metal atom is, by
 > itself, all you need to know; there are almost never two, so you're 
wasting
 > 15 bits if you encode metals the same way you encode nitro groups 
or methyls.
 > Using 16 bits to encode every functional group is simple, but can 
waste a lot
 > of space, which translates to slower searching.
> 
>> Yes, this would lead towards something like the CACTVS keys in PubChem. 
>> For OB, the significance of each pattern could be coded in the input 
>> patterns file and control the bit-setter.

Following this discussion, I have had a go at modifying finger3.cpp so 
that PatternFP is more versatile, while remaining backward 
compatibility - FP3 and FP4 are the same as they were. I haven't been 
able to see any other established schemes (like the CACTVS one 
mentioned above) so maybe there are things that could be improved.

Each pattern in a datafile like SMARTS_InteLigand.txt can now have two 
optional parameters, one specifying a number of occurrences (m) and 
the other the number of bits (n) for that pattern.

If n=1 (the default) or absent then the pattern will match only when 
there are greater than m matches in the molecule.

If m=0 (the default) then the pattern has n bits (and extra weight in 
any similarity tests).

If the parameters are n-1 and n, a bit is set for each of the 
conditions, number of matches >=n, >=n-1, ... , >=1   This can be used 
to distinguish structures with many similar atoms like n-alkanes.
So a datafile with just one line:
Secondary_carbon: [CX4H2]([#6])[#6] 13 14
will distinguish all n-alkanes up to C14H30. But the parameters can 
have any positive value and other behaviours are possible.

As with some other plugin classes, fingerprints based on Pattern FP 
can now be defined without re-compiling, by making an entry in 
plugindefines.txt. This looks like

PatternFP
MACCS          #ID of this fingerprint type
MACCS.txt      #File containing the SMARTS patterns

The data for this fingerprint is taken from RDKit and is in yet 
another file format. (Andrew Dalke won't be pleased.) This has about 
160 patterns, but I'm not sure about how complete it is. It makes use 
of the occurrences parameter above.

The new and modified fiiles are in the trunk on SVN.

Chris