Re: [Rdkit-discuss] identify isomers using canonical SMILES

[Andrew D. <da...@da...>]

On Feb 23, 2009, at 4:48 PM, Greg Landrum wrote:
> Somehow it has happened that the RDKit does not have a function to
> generate the chemical formula for a molecule, so one would need to
> write it from scratch. Here's a simple (and relatively untested) way
> of doing this:

Ideally it would generate the Hill formula,

>   ks = cnts.keys()
>   ks.sort()
>   res=''
>   for k in ks:
>     res+=k
>     if cnts[k]>1:
>       res+=str(cnts[k])

would be more like:

   ks = cnts.keys()
   # Alphabetize everything
   ks.sort()

   # Put into Hill order (C then H then everything else)
   if "C" in cnts:
     ks.remove("C")
     ks.insert(0, "C")
     if "H" in cnts:
       ks.remove("H")
       ks.insert(1, "H")

   ...

There are other solutions which are more efficient for large N, but  
there's only about 100 elements in the world, and only a handful in  
most compounds.


				Andrew
				da...@da...